[Mesa-dev] [PATCH] i965: Set SURFACE_STATE vertical alignment bit on Ivybridge.

[Ian Romanick]

On 11/23/2011 06:34 PM, Kenneth Graunke wrote:
> See intel_vertical_texture_alignment_unit() in intel_tex_layout.c;
> certain surface types require setting this to VALIGN_4.
>
> Analogous to commit dd0e46c4102976b7d317104ecd1bb565ac34613a on Gen6.
>
> Fixes piglit test fbo-generatemipmap-formats with the
> GL_ARB_depth_texture and GL_EXT_packed_depth_stencil arguments.
>
> Signed-off-by: Kenneth Graunke<kenneth at whitecape.org>
> ---
>   src/mesa/drivers/dri/i965/gen7_wm_surface_state.c |    7 +++++++
>   1 files changed, 7 insertions(+), 0 deletions(-)
>
> v2: Do what Chad said
>
> Good call.  No difference in piglit compared to the first, but a good idea
> nonetheless.
>
> diff --git a/src/mesa/drivers/dri/i965/gen7_wm_surface_state.c b/src/mesa/drivers/dri/i965/gen7_wm_surface_state.c
> index ec72c1e..d20bdb5 100644
> --- a/src/mesa/drivers/dri/i965/gen7_wm_surface_state.c
> +++ b/src/mesa/drivers/dri/i965/gen7_wm_surface_state.c
> @@ -59,6 +59,7 @@ gen7_update_texture_surface(struct gl_context *ctx, GLuint unit)
>      struct brw_context *brw = brw_context(ctx);
>      struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current;
>      struct intel_texture_object *intelObj = intel_texture_object(tObj);
> +   struct intel_mipmap_tree *mt = intelObj->mt;
>      struct gl_texture_image *firstImage = tObj->Image[0][tObj->BaseLevel];
>      struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit);
>      const GLuint surf_index = SURF_INDEX_TEXTURE(unit);
> @@ -71,6 +72,9 @@ gen7_update_texture_surface(struct gl_context *ctx, GLuint unit)
>   			  sizeof(*surf), 32,&brw->bind.surf_offset[surf_index]);
>      memset(surf, 0, sizeof(*surf));
>
> +   if (mt->align_h == 4)
> +      surf->ss0.vertical_alignment = 1;
> +
>      surf->ss0.surface_type = translate_tex_target(tObj->Target);
>      surf->ss0.surface_format = translate_tex_format(firstImage->TexFormat,
>                                                      firstImage->InternalFormat,
> @@ -200,6 +204,9 @@ gen7_update_renderbuffer_surface(struct brw_context *brw,
>   			  sizeof(*surf), 32,&brw->bind.surf_offset[unit]);
>      memset(surf, 0, sizeof(*surf));
>
> +   if (irb->mt->align_h == 4)
> +      surf->ss0.vertical_alignment = 1;
> +

Does adding this cause more tests to pass compared to the previous 
version of this patch?  If not, can we concoct a test that does change?

>      switch (irb->Base.Format) {
>      case MESA_FORMAT_SARGB8:
>         /* without GL_EXT_framebuffer_sRGB we shouldn't bind sRGB