[Mesa-dev] [PATCH 2/3] nv50: add more RGB10A2 formats

[Ilia Mirkin]

On Wed, Jan 15, 2014 at 6:37 AM, Ilia Mirkin <imirkin at alum.mit.edu> wrote:
> On Wed, Dec 25, 2013 at 11:53 AM, Christoph Bumiller
> <e0425955 at student.tuwien.ac.at> wrote:
>> ---
>>  src/gallium/drivers/nouveau/nv50/nv50_formats.c | 7 +++++++
>>  1 file changed, 7 insertions(+)
>>
>> diff --git a/src/gallium/drivers/nouveau/nv50/nv50_formats.c b/src/gallium/drivers/nouveau/nv50/nv50_formats.c
>> index 0a7e812..b301890 100644
>> --- a/src/gallium/drivers/nouveau/nv50/nv50_formats.c
>> +++ b/src/gallium/drivers/nouveau/nv50/nv50_formats.c
>> @@ -202,6 +202,8 @@ const struct nv50_format nv50_format_table[PIPE_FORMAT_COUNT] =
>>         TBV, 1),
>>     C4A(R10G10B10A2_SNORM, NONE, C0, C1, C2, C3, SNORM, 10_10_10_2, TV, 0),
>>     C4A(B10G10R10A2_SNORM, NONE, C2, C1, C0, C3, SNORM, 10_10_10_2, TV, 1),
>> +   C4A(R10G10B10A2_UINT, RGB10_A2_UINT, C0, C1, C2, C3, UINT, 10_10_10_2, TRV, 0),
>> +   C4A(B10G10R10A2_UINT, RGB10_A2_UINT, C2, C1, C0, C3, UINT, 10_10_10_2, TV, 0),
>>
>>     F3B(R11G11B10_FLOAT, R11G11B10_FLOAT, C0, C1, C2, xx, FLOAT, 11_11_10, IB),
>>
>> @@ -394,6 +396,11 @@ const struct nv50_format nv50_format_table[PIPE_FORMAT_COUNT] =
>>     F1A(R16_SSCALED, NONE, C0, xx, xx, xx, SSCALED, 16, V),
>>     F1A(R16_USCALED, NONE, C0, xx, xx, xx, USCALED, 16, V),
>>
>> +   C4A(R10G10B10A2_USCALED, NONE, C0, C1, C2, C3, USCALED, 10_10_10_2, V, 0),
>> +   C4A(R10G10B10A2_SSCALED, NONE, C0, C1, C2, C3, SSCALED, 10_10_10_2, V, 0),
>> +   C4A(B10G10R10A2_USCALED, NONE, C0, C1, C2, C3, USCALED, 10_10_10_2, V, 1),
>> +   C4A(B10G10R10A2_SSCALED, NONE, C0, C1, C2, C3, SSCALED, 10_10_10_2, V, 1),
>> +
>>     C4A(R8G8B8A8_SSCALED, NONE, C0, C1, C2, C3, SSCALED, 8_8_8_8, V, 0),
>>     C4A(R8G8B8A8_USCALED, NONE, C0, C1, C2, C3, USCALED, 8_8_8_8, V, 0),
>>     F3A(R8G8B8_UNORM, NONE, C0, C1, C2, xx, UNORM, 8_8_8, V),
>
> FWIW I ran piglit on this (with your other patches that cause the
> RGB10A2 extensions to actually be turned on), and I saw this failure:
>
> $ bin/draw-vertices-2101010 -auto
> Int vertices - 2/10/10/10
> Unsigned Int vertices - 2/10/10/10
> Int Color - 2/10/10/10
> Probe color at (45,5)
>   Expected: 1.000000 0.000000 0.000000 0.333000
>   Observed: 1.000000 0.000000 0.000000 0.000000
> Unsigned Int Color - 2/10/10/10
> Int BGRA Color - 2/10/10/10
> Probe color at (85,5)
>   Expected: 0.000000 0.000000 1.000000 0.333000
>   Observed: 0.000000 0.000000 1.000000 0.000000
> Unsigned Int BGRA Color - 2/10/10/10
> Int 2/10/10/10 - test ABI
> Unsigned 2/10/10/10 - test ABI

BTW, I tried to look into this...

(a) Shouldn't the second C4A in the first hunk have a 1 for the BGRA
bit? And similarly, shouldn't the components be flipped for the BGRA
formats in the second hunk? Doing this made the
GL_ARB_texture_rgb10_a2ui tests pass (well, mostly -- projective mode,
whatever that is, still doesn't work, but that wasn't tested before)

(b) I think that the draw-vertices test might be broken, but not in
any way that matters:
  - its i32to2 function is wrong. but it's fine for 0 which is what's
being passed in
  - 0 for a signed (2's complement) A2 component should become
0.66666, not 0.33333. I think. It's the third of 4 values -- -2, -1,
0, 1. But that's not what we're seeing anyways.

(c) When I ran on the blob earlier, it had identical failures, and so
does llvmpipe.

>
> There were also ARB_texture_rgb10_a2ui failures with texwrap
> bordercolor's, but there are lots of other such failures with integer
> formats, so I assume that's all part of the same thing. For reference:
>
> $ bin/texwrap GL_ARB_texture_rgb10_a2ui bordercolor -fbo -auto
> Testing GL_ARB_texture_rgb10_a2ui.
> Testing the border color only.
> Testing GL_RGB10_A2UI, border color only
> Fail with NEAREST and CLAMP_TO_BORDER at (95,36) @ 0,0
>   Expected: 25 229 127 170
>   Observed: 0 0 0 0
> Fail with NEAREST and MIRROR_CLAMP_TO_BORDER_EXT at (273,36) @ 0,0
>   Expected: 25 229 127 170
>   Observed: 0 0 0 0
>
> $ bin/texwrap GL_ARB_texture_rgb10_a2ui bordercolor swizzled -fbo -auto
> Testing GL_ARB_texture_rgb10_a2ui.
> Testing the border color only.
> Using texture swizzling.
> Testing GL_RGB10_A2UI, swizzled, border color only
> Fail with NEAREST and CLAMP_TO_BORDER at (95,36) @ 0,0
>   Expected: 127 25 229 170
>   Observed: 0 0 0 0
> Fail with NEAREST and MIRROR_CLAMP_TO_BORDER_EXT at (273,36) @ 0,0
>   Expected: 127 25 229 170
>   Observed: 0 0 0 0