Okay I think this is good enough for now, I can't see any other reason
for mesa to want to use a sampler view so lets just leave it at all the A->X conversions for now.
I've been running gnome-shell under r300g with this for day or so and it seems fine.
Signed-off-by: Dave Airlie <airlied@redhat.com>
This passes on r300g, the only bit I'm not really sure about is the handling
of the sampler_view in st_atom_texture.c, I unreference it there if the swizzle
value changes and I also have to create a new set of functions to create a new
one since the u_sampler.c ones don't handle swizzle so much.
adds r300g + softpipe enables, I think other drivers could pass easily enough.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Previously, when we created a gallium texture for a corresponding Mesa
texture we'd only allocate space for mipmap levels >= BaseLevel.
This patch undoes that mechanism. This fixes a render-to-texture bug
when rendering to level 0 when BaseLevel=1.
Also, it makes sense to allocate the whole texture object memory when
BaseLevel > 0 since a common use of GL_TEXTURE_BASE_LEVEL is to
progressively load/render mipmaps. Eventually, the app almost always
fills in the level=0 mipmap image.
Finally, the texture image code is bit easier to understand now.
Before, we were sometimes rendering into a stale texture because
st_finalize_texture() would discard the old texture and create a new one.
Moved st_update_framebuffer atom after texture validation so that we
can create a new renderbuffer surface if the texture changes.
Also, split texture validation into two parts: finalize_textures and
update_textures. Do finalize_textures first to avoid getting into the
situtation where we're doing a pipe->surface_copy() mid-way through
state validation.
Some debug code still in place, but disabled...
If we run out of texture memory we may not have the texture needed by the
fragment shader. If this happens, plug in a color passthrough shader.
So instead of crashing, we just don't see the texture.
GL_OUT_OF_MEMORY is raised, of course.
Calling glTexSubImage() or glTexImage() to replace texture data didn't
reliably cause pipe->set_sampler_texture() to get called so drivers didn't
always get notified of new texture data.
The st_texture_object->pt pointer doesn't always indicate changed data so
added a dirtyData field.
