On Wed, Oct 01, 2014 at 03:23:44PM +0200, Stefan Wahren wrote:

Is it better to let the core handles the ramp delay instead of set_voltage_sel
with a busy wait?

I've found that polling DC_OK is only reliable for increasing voltages. So i
think about defining ramp delay in the regulator description.

It's better to use the notification from the device if it is reliable;
dead reckoning isn't going to work well either for voltage reduction
since the time taken to implement a drop in voltage is more a function
of the loading than anything else, the regulator can't really control
it.  For most DC-DC loads it'll be quick enough though.