[Intel-gfx] [PATCH] drm/i915: Fix Limited Range Color Handling

[Shankar, Uma]

>-----Original Message-----
>From: Ville Syrjälä [mailto:ville.syrjala at linux.intel.com]
>Sent: Friday, December 22, 2017 9:43 PM
>To: Shankar, Uma <uma.shankar at intel.com>
>Cc: intel-gfx at lists.freedesktop.org; Lin, Johnson <johnson.lin at intel.com>;
>Syrjala, Ville <ville.syrjala at intel.com>; Lankhorst, Maarten
><maarten.lankhorst at intel.com>
>Subject: Re: [Intel-gfx] [PATCH] drm/i915: Fix Limited Range Color Handling
>
>On Fri, Dec 22, 2017 at 05:54:31PM +0200, Ville Syrjälä wrote:
>> On Fri, Dec 22, 2017 at 08:34:47PM +0530, Uma Shankar wrote:
>> > From: Johnson Lin <johnson.lin at intel.com>
>> >
>> > Some panels support limited range output (16-235) compared to full
>> > range RGB values (0-255). Also userspace can control the RGB range
>> > using "Broadcast RGB" property. Currently the code to handle full
>> > range to limited range is broken. This patch fixes the same by
>> > properly scaling down all the full range co-efficients with limited
>> > range scaling factor.
>> >
>> > Signed-off-by: Johnson Lin <johnson.lin at intel.com>
>> > Signed-off-by: Uma Shankar <uma.shankar at intel.com>
>> > ---
>> >  drivers/gpu/drm/i915/intel_color.c |   16 ++++++++++++----
>> >  1 file changed, 12 insertions(+), 4 deletions(-)
>> >
>> > diff --git a/drivers/gpu/drm/i915/intel_color.c
>> > b/drivers/gpu/drm/i915/intel_color.c
>> > index aa66e95..777ce26 100644
>> > --- a/drivers/gpu/drm/i915/intel_color.c
>> > +++ b/drivers/gpu/drm/i915/intel_color.c
>> > @@ -94,16 +94,24 @@ static void ctm_mult_by_limited(uint64_t *result,
>int64_t *input)
>> >  	for (i = 0; i < 9; i++)
>> >  		result[i] = 0;
>>
>> This can go.
>>
>> >
>> > -	for (i = 0; i < 3; i++) {
>> > -		int64_t user_coeff = input[i * 3 + i];
>> > +	for (i = 0; i < 9; i++) {
>> > +		int64_t user_coeff = input[i];
>>
>> That's not a two's complement number so should probably be u64.
>>
>> >  		uint64_t limited_coeff = CTM_COEFF_LIMITED_RANGE >> 2;
>> >  		uint64_t abs_coeff = clamp_val(CTM_COEFF_ABS(user_coeff),
>> >  					       0,
>> >  					       CTM_COEFF_4_0 - 1) >> 2;
>>
>> Seems to me that we should be able to drop the >>2 from the
>> limited_coeff since it's always < 1.0, make both of these u32 and then
>> use
>> 'mul_u32_u32() >> 30' to get the answer.
>>
>> >
>> > -		result[i * 3 + i] = (limited_coeff * abs_coeff) >> 27;
>> > +		/*
>> > +		 * By scaling every co-efficient with limited range (235-16)
>> > +		 * vs full range (0-255) the final o/p will be scaled down to
>> > +		 * fit in the limited range supported by the panel.
>> > +		 * Need to shift the multiplication result by 28 as the floating
>> > +		 * count expected is of 32bits, multiplication here is done in
>> > +		 * U2.30 so result need to be right shifted by 60-32 = 28
>> > +		 */
>> > +		result[i] = (limited_coeff * abs_coeff) >> 28;
>> >  		if (CTM_COEFF_NEGATIVE(user_coeff))
>>
>> And this can be replaced with just
>>
>> result[i] |= user_coeff & CTM_COEFF_SIGN;
>>
>> to eliminate the silly branch.
>
>I wonder if we could also get some igts for this limited range stuff.
>Maybe something like:
>1. grab crc from black/white screen with limited range 2. grab cdc from almost
>black/white screen with full range 3. compare the two
>
>Repeat with an identity csc matrix set.
>
>Not sure if we can get the crcs to match in these two cases though.
>Hard to say without trying it out.
>

Hi Ville,
We will try this approach and update. Ideally crc's should match, will confirm the
behavior on hardware.

Thanks & Regards,
Uma Shankar

>>
>>
>> > -			result[i * 3 + i] |= CTM_COEFF_SIGN;
>> > +			result[i] |= CTM_COEFF_SIGN;
>> >  	}
>> >  }
>> >
>> > --
>> > 1.7.9.5
>> >
>> > _______________________________________________
>> > Intel-gfx mailing list
>> > Intel-gfx at lists.freedesktop.org
>> > https://lists.freedesktop.org/mailman/listinfo/intel-gfx
>>
>> --
>> Ville Syrjälä
>> Intel OTC
>
>--
>Ville Syrjälä
>Intel OTC