Re: [PATCH v2 2/2] dt-bindings: hwmon: Add sample averaging property for ADM1275

[Guenter Roeck <linux@roeck-us.net>]

On 2/24/22 23:31, Krzysztof Kozlowski wrote:
> On 25/02/2022 08:06, Krzysztof Kozlowski wrote:
>> On 24/02/2022 16:43, Potin Lai wrote:
>>> Add new properties for binding sample averaging in PMON_CONFIG register
>>>
>>> - adi,volt-curr-sample-average
>>> - adi,power-sample-average
>>>
>>> Signed-off-by: Potin Lai <potin.lai@quantatw.com>
>>> ---
>>>   .../bindings/hwmon/adi,adm1275.yaml           | 44 +++++++++++++++++++
>>>   1 file changed, 44 insertions(+)
>>>
>>> diff --git a/Documentation/devicetree/bindings/hwmon/adi,adm1275.yaml b/Documentation/devicetree/bindings/hwmon/adi,adm1275.yaml
>>> index 223393d7cafd..325f6827648f 100644
>>> --- a/Documentation/devicetree/bindings/hwmon/adi,adm1275.yaml
>>> +++ b/Documentation/devicetree/bindings/hwmon/adi,adm1275.yaml
>>> @@ -37,6 +37,48 @@ properties:
>>>       description:
>>>         Shunt resistor value in micro-Ohm.
>>>   
>>> +  adi,volt-curr-sample-average:
>>> +    description: |
>>> +      A value to configure VI_AVG in PMON_CONFIG register to indicate a
>>> +      number of samples to be used to report voltage and currentvalues.
>>
>> missing space after current.
>>
>>> +      If set to 7, the 128 samples averaging would be used.
>>> +
>>> +    $ref: /schemas/types.yaml#/definitions/uint8
>>
>> Make it a uint32.
>>
>> The previous usage of this field was more appropriate. Instead of
>> keeping register values in DT, it's better to keep logical value. What
>> if in next cheap the register values have calculation method?
>>
>> This should be like in v1 - enum for number of samples to take in averaging.
>>
> 
> One more thought: this field could also stay in current approach if you
> change the meaning from "value to configure VI_AVG" to something like
> "the exponent used to determine the number of samples, where the base is 2".
> 
> This approach would allow you to skip the "ilog" in the code. It sill
> won't be that easily scalable if another chip comes with different
> formula, but I think that's unlikely.
> 

The standard hwmon ABI expects the number of samples, it isn't always a
power of 2, and the driver already implements it (with ilog2) as sysfs
attribute. I don't really see the point of "optimizing" something
like this to be chip specific just to avoid some error checking.

Guenter