[Mesa-dev] [PATCH] fix minor error in YUV2RGB matrix used in shader

[Lin, Johnson]

Yes, We can.  Will do it

-----Original Message-----
From: Eric Anholt [mailto:eric at anholt.net] 
Sent: Saturday, April 29, 2017 6:02 AM
To: Lin, Johnson <johnson.lin at intel.com>; mesa-dev at lists.freedesktop.org
Cc: Lin, Johnson <johnson.lin at intel.com>
Subject: Re: [Mesa-dev] [PATCH] fix minor error in YUV2RGB matrix used in shader

Johnson Lin <johnson.lin at intel.com> writes:

> The matrix used for YCbCr to RGB is listed in Wiki 
> https://en.wikipedia.org/wiki/YCbCr;
> There is minor error in the matrix constant: 0.0625=16/256 should be 
> 16.0/255,  and 0.5=128.0/256 should be 128.0/255.
> Note that conversion from a 0-255 byte number to 0-1.0 float is to 
> divide by 255  instead of 256. That's we get 255=1.0f.
> By the constant change we can see the CSC result is bit aligned with 
> Wiki conversion result and FFMPeg result.
> Otherwise in some situation, there will be one bit difference
>
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100854
> ---
>  src/compiler/nir/nir_lower_tex.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/src/compiler/nir/nir_lower_tex.c 
> b/src/compiler/nir/nir_lower_tex.c
> index 352d1499bc8d..385739a56a71 100644
> --- a/src/compiler/nir/nir_lower_tex.c
> +++ b/src/compiler/nir/nir_lower_tex.c
> @@ -244,9 +244,9 @@ convert_yuv_to_rgb(nir_builder *b, nir_tex_instr *tex,
>     nir_ssa_def *yuv =
>        nir_vec4(b,
>                 nir_fmul(b, nir_imm_float(b, 1.16438356f),
> -                        nir_fadd(b, y, nir_imm_float(b, -0.0625f))),
> -               nir_channel(b, nir_fadd(b, u, nir_imm_float(b, -0.5f)), 0),
> -               nir_channel(b, nir_fadd(b, v, nir_imm_float(b, -0.5f)), 0),
> +                        nir_fadd(b, y, nir_imm_float(b, -0.0627451f))),
> +               nir_channel(b, nir_fadd(b, u, nir_imm_float(b, -0.50196078431f)), 0),
> +               nir_channel(b, nir_fadd(b, v, nir_imm_float(b, 
> + -0.50196078431f)), 0),
>                 nir_imm_float(b, 0.0));

Could we use 16.0/255.0 and 128.0/255.0, instead of magic-looking numbers?  With that, it will be:

Reviewed-by: Eric Anholt <eric at anholt.net>