[Intel-gfx] [PATCH] drm/i915/icl: Handle YCbCr to RGB conversion for BT2020 case

[Ville Syrjälä]

On Thu, May 09, 2019 at 02:54:19PM +0000, Shankar, Uma wrote:
> 
> 
> >-----Original Message-----
> >From: Ville Syrjälä [mailto:ville.syrjala at linux.intel.com]
> >Sent: Tuesday, May 7, 2019 9:08 PM
> >To: Shankar, Uma <uma.shankar at intel.com>
> >Cc: Sharma, Shashank <shashank.sharma at intel.com>; intel-
> >gfx at lists.freedesktop.org
> >Subject: Re: [Intel-gfx] [PATCH] drm/i915/icl: Handle YCbCr to RGB conversion for
> >BT2020 case
> >
> >On Tue, May 07, 2019 at 02:35:15PM +0000, Shankar, Uma wrote:
> >>
> >>
> >> >-----Original Message-----
> >> >From: Intel-gfx [mailto:intel-gfx-bounces at lists.freedesktop.org] On
> >> >Behalf Of Ville Syrjälä
> >> >Sent: Tuesday, May 7, 2019 7:37 PM
> >> >To: Sharma, Shashank <shashank.sharma at intel.com>
> >> >Cc: intel-gfx at lists.freedesktop.org
> >> >Subject: Re: [Intel-gfx] [PATCH] drm/i915/icl: Handle YCbCr to RGB
> >> >conversion for
> >> >BT2020 case
> >> >
> >> >On Tue, May 07, 2019 at 06:32:57PM +0530, Shashank Sharma wrote:
> >> >> From: Uma Shankar <uma.shankar at intel.com>
> >> >>
> >> >> Currently input csc for YCbCR to RGB conversion handles only
> >> >> BT601 and Bt709. Extending it to support BT2020 as well.
> >> >>
> >> >> Signed-off-by: Uma Shankar <uma.shankar at intel.com>
> >> >> Signed-off-by: Shashank Sharma <shashank.sharma at intel.com>
> >> >> ---
> >> >>  drivers/gpu/drm/i915/intel_sprite.c | 24 ++++++++++++++++++++++++
> >> >>  1 file changed, 24 insertions(+)
> >> >>
> >> >> diff --git a/drivers/gpu/drm/i915/intel_sprite.c
> >> >> b/drivers/gpu/drm/i915/intel_sprite.c
> >> >> index 44aaeac1b2ed..2536e757bec2 100644
> >> >> --- a/drivers/gpu/drm/i915/intel_sprite.c
> >> >> +++ b/drivers/gpu/drm/i915/intel_sprite.c
> >> >> @@ -433,6 +433,18 @@ icl_program_input_csc(struct intel_plane *plane,
> >> >>  			0x9EF8, 0x7800, 0xABF8,
> >> >>  			0x0, 0x7800,  0x7ED8,
> >> >>  		},
> >> >> +		/*
> >> >> +		 * BT.2020 full range YCbCr -> full range RGB
> >> >> +		 * The matrix required is :
> >> >> +		 * [1.000, 0.000, 1.474,
> >> >> +		 *  1.000, -0.1645, -0.5713,
> >> >> +		 *  1.000, 1.8814, 0.0000]
> >> >> +		 */
> >> >> +		[DRM_COLOR_YCBCR_BT2020] = {
> >> >> +			0x7BC8, 0x7800, 0x0,
> >> >> +			0x8928, 0x7800, 0xAA88,
> >> >> +			0x0, 0x7800, 0x7F10,
> >> >> +		},
> >> >>  	};
> >> >>
> >> >>  	/* Matrix for Limited Range to Full Range Conversion */ @@ -461,6
> >> >> +473,18 @@ icl_program_input_csc(struct intel_plane *plane,
> >> >>  			0x8888, 0x7918, 0xADA8,
> >> >>  			0x0, 0x7918,  0x6870,
> >> >>  		},
> >> >> +		/*
> >> >> +		 * BT.2020 Limited range YCbCr -> full range RGB
> >> >> +		 * The matrix required is :
> >> >> +		 * [1.164, 0.000, 1.717,
> >> >> +		 *  1.138, -0.1873, -0.6504,
> >> >> +		 *  1.1380, 2.1417, 0.0000]
> >> >
> >> >Where are those 1.138 coming from?
> >>
> >> Hi Ville,
> >> This is the original YCBCR to RGB BT2020 matrix:
> >> {
> >>         1.00000000000,  0.00000000000,  1.47460000000,
> >>         1.00000000000,  -0.16455312684, -0.57135312684,
> >>         1.00000000000,  1.88140000000,  0.00000000000 };
> >>
> >> We have to convert Limited Range YCbCr to Full Range RGB. Hence we need to
> >apply a scale factor:
> >>  yscalefactor = 219.0 * normalizingfactor;  cbcrscalefactor = 224.0 *
> >> normalizingfactor;
> >>
> >>  /* Scale factors are inverted for LR to FR conversion */
> >> yscalefactor = 1.0 / yscalefactor;  cbcrscalefactor = 1.0 /
> >> cbcrscalefactor;
> >>
> >> This yields the above results.
> >
> >Those are the coefficients for Y, so they should still be the same for all three output
> >channels.
> >
> >igt_color_encoding gives me:
> >|1.1644, 0.0000, 1.6787,|
> >|1.1644,-0.1873,-0.6504,|
> >|1.1644, 2.1418, 0.0000,|
> 
> Ok, I used the igt_color_encoding method and able to get values what you got.
> Will update the matrix. Thanks Ville.
> 
> >Looks like we're also misprogramming the Y pre-offset for the full range YCbCr case.
> For full range, I am getting same values as programmed above. Looks ok, can you double check.

The matrix itself looks OK (some minor rounding differences perhaps, but
nothing major). But the Y preoffset should be zero for full range
YCbCr.

-- 
Ville Syrjälä
Intel