Re: [PATCH 2/2] drm/i915: respect the VBT minimum backlight brightness

[Jesse Barnes <jesse.barnes@intel.com>]

On Sat, 28 Jun 2014 16:45:03 +0300
Jani Nikula <jani.nikula@intel.com> wrote:
> >> +/* Scale user_level in range [0..user_max] to [0..hw_max], clamping the result
> >> + * to [hw_min..hw_max]. */
> >> +static inline u32 clamp_user_to_hw(struct intel_connector *connector,
> >> +				   u32 user_level, u32 user_max)
> >> +{
> >> +	struct intel_panel *panel = &connector->panel;
> >> +	u32 hw_level;
> >> +
> >> +	hw_level = scale(user_level, 0, user_max, 0, panel->backlight.max);
> >> +	hw_level = clamp(hw_level, panel->backlight.min, panel->backlight.max);
> >> +
> >> +	return hw_level;
> >> +}
> >
> > I like the direction here, but does this mean some user values will
> > potentially be no-ops?  E.g. the low 10 values or something would all
> > map to backlight.min depending on the min?
> 
> This patch doesn't really change the fact. For a max backlight of, say,
> 100000, we're bound to not have user perceivable difference between
> consecutive levels. I agree we should have a fixed, limited range here.
> 
> This also depends on the backlight modulation frequency, see my earlier
> message about this: http://mid.gmane.org/87iooecg1e.fsf@intel.com
> 
> > I just want to make sure every value we expose to userspace is
> > meaningful, and somehow equidistant from the values next to it, so we
> > have nice, smooth backlight transitions, and fades look good (on that
> > front, is 256 enough?  or should we have a scaled range up to 1024 or
> > so?)
> 
> Probably even fewer than that is enough.
> 
> But you bring up another requirement, equidistant - do you mean in terms
> of luminance? Almost certainly a linear mapping to the duty cycle is not
> going to give you a linear luminance! It's possible to define a curve
> for this in the acpi opregion; patches to implement this are welcome. ;)

Ok, I guess both of these are projects for future patches.  For making
the brightness levels evenly spaced, yeah I guess we'd need to apply a
rough approximation of a luminosity function (doesn't look like a
simple exponential curve, oh well).

Jesse