[Mesa-dev] [PATCH v6 0/5] improved the support for ETC2 formats on Gen 7

[Nanley Chery]

On Fri, Feb 15, 2019 at 03:29:39PM +0200, Eleni Maria Stea wrote:
> Intel Gen7 GPUs don't support the ETC2 formats natively and in order to
> show the pixels properly we decompress them and create decompressed
> miptrees. The problem with that is that the functions that map the
> miptrees for reading (for example the GetCompressed* calls), and would
> be supposed to read compressed pixel values, would read decompressed
> values instead unless if we prevented this with assertions that make
> the user programs either crash or misfunction.
> 
> These patches are an attempt to give a solution to this problem by using 2
> miptrees: the main to store the ETC values and the generic shadow
> (mt->shadow) to store the decompressed values. Each time that the main
> miptree is mapped for writing we set a flag that the shadow will need
> update and we check this flag before every draw call to update the
> shadow miptree. (We perform the check right before drawing to avoid
> missing changes from functions like the CopyImageSubData in the next 
> frame). Then we map the shadow for sampling. This way, we can render the
> images using the decompressed pixels of the shadow but we return the
> compressed ones from the main when the texture is mapped for reading.
> 
> Also, the OES_copy_image extension that couldn't work on Gen 7 due to the
> lack of the ETC support is now enabled back.
> 
> Finally, the following glcts and piglit tests pass:
> 
> On HSW (previously failing):
> ----------------------------
> KHR-GL46.direct_state_access.textures_compressed_subimage
> 
> On HSW and IVB (previously skipped):
> -------------------------------------
> dEQP-GLES31.functional.texture.border_clamp.formats.compressed_srgb8_alpha8_etc2_eac.*
>    (6 tests)
> dEQP-GLES31.functional.texture.border_clamp.formats.compressed_srgb8_etc2.*
>    (6 tests)
> dEQP-GLES31.functional.texture.border_clamp.formats.compressed_srgb8_punchthrough_alpha1_etc2.*
>    (6 tests)
> 
> On HSW, IVB, SNB (previously skipped):
> ---------------------------------------
> dEQP-GLES3.functional.texture.format.compressed.*
>    (12 tests)
> dEQP-GLES3.functional.texture.wrap.etc2_eac_srgb8_alpha8.*
>    (36 tests)
> dEQP-GLES3.functional.texture.wrap.etc2_srgb8.*
>    (36 tests)
> dEQP-GLES3.functional.texture.wrap.etc2_srgb8_punchthrough_alpha1.*
>    (36 tests)
> 
> piglit.spec.!opengl es 3_0.oes_compressed_etc2_texture-miptree_gles3 
>    (srgb8, srgb8-alpha, srgb8-punchthrough-alpha1)
> piglit.spec.arb_es3_compatibility.oes_compressed_etc2_texture-miptree
>    (srgb8 compat, srgb8 core, srgb8-alpha8 compat, srgb8-alpha8 core,
>     srgb8-punchthrough-alpha1 compat, srgb8-punchthrough-alpha1 core)
> (9 tests)
> 
> Total tests passing: 148
> 
> Eleni Maria Stea (4):
>   i965: Faking the ETC2 compression on Gen < 8 GPUs using two miptrees.
>   i965: Fixed the CopyImageSubData for ETC2 on Gen < 8
>   i965: Enabled the OES_copy_image extension on Gen 7 GPUs
>   i965: Removed the field etc_format from the struct intel_mipmap_tree
> 

These patches are
Reviewed-by: Nanley Chery <nanley.g.chery at intel.com>

I like how this series turned out. Thank you!

> Nanley Chery (1):
>   i965: Rename intel_mipmap_tree::r8stencil_* -> ::shadow_*
> 
>  src/mesa/drivers/dri/i965/brw_draw.c          |   5 +
>  .../drivers/dri/i965/brw_wm_surface_state.c   |  15 +-
>  src/mesa/drivers/dri/i965/intel_extensions.c  |  16 +-
>  src/mesa/drivers/dri/i965/intel_mipmap_tree.c | 170 ++++++++++--------
>  src/mesa/drivers/dri/i965/intel_mipmap_tree.h |  48 +++--
>  5 files changed, 149 insertions(+), 105 deletions(-)
> 
> -- 
> 2.20.1
>