07:33 <wens> in the regulator driver helpers, the voltage selectors are down-shifted so the lowest bit is at offset 0; we don't do that for the range selectors though, and end up with "0x0, 0x0, 0x0, 0x40, 0x80, 0x80, 0x80" for them in bd718x7-regulator.c

07:34 <wens> I find it a bit confusing, and harder to match driver values against datasheets, which tend to list values in the bitfield without the extra shift

07:35 <wens> I'm wondering why it was done this way. I couldn't find any examples of vsel_range_mask being non-contiguous upstream.

10:46 <mvaittin> wens: You're a bit late with this question :) I might have been able to answer this 4 or 5 years ago ^_^; I have some very hazy feeling that there was some "inbuilt assumption" about selectors when pickable ranges are used.

10:49 <mvaittin> wens: Oh, or are you wondering why the voltage selectors are shifted? This decision was done before I started with the regulator stuff

11:26 <mvaittin> wens: I did some necromancy and it seems that my faint memory about "inbuilt expectation" is explained in comment at regulator_list_voltage_pickable_linear_range(). I think this is not related to your question though. So, would you suggest that the range selectors would start from 0 and be shifted using FIELD_PREP() equivalent before writing to HW?

11:30 <mvaittin> (So, for the bd71837_buck5_volt_range_sel it would be "0x0, 0x0, 0x0, 0x1, 0x2, 0x2, 0x2"?) - For my eyes this is not really clearer - but this is just my limited experience and probably the reason why pickable ranges look like they do look like :)

13:21 <wens> mvaittin: "start from 0 and be shifted using FIELD_PREP()", basically yes.

13:24 <wens> I mostly found it odd, since vsel starts from bit 0, I assumed vsel_range started from bit 0 as well

13:25 <wens> about readability, I guess it depends on how the datasheets are written.

13:27 <wens> I'm working on PMICs that, in addition to the standard voltage selector (whose values has holes), have a "output calibration" field that can do 0.00v ~ 0.1v adjustments in 0.01v steps. so I'm putting the voltage selector in vsel_range, and the calibration field in vsel.

13:28 <wens> the voltage selector is written in 0bxxxx style, even though it's field is GENMASK(11, 8).

13:28 <wens> not having to left shift the values manually when writing the driver is mostly a convenience.

13:55 <broonie> I'd generally expect anything going into a bitfield to be written as numbers written into that field (ie, the value will need to be shifted if the field is not at bit 0).

13:58 <broonie> That looks like an oversight with the range selectors, probably due to the handling being split over multiple functions.

20:20 <marex> linusw__: so hum, sandisk card works, another Goodram card also works, the kingston card does not work