[Mesa-dev] [PATCH 1/2] mesa: support internalFormat=GL_BGRA in TexImage2D

[Brian Paul]

2011/1/24 Kristian Høgsberg <krh at bitplanet.net>:
> 2011/1/24 Brian Paul <brianp at vmware.com>:
>> On 01/24/2011 04:18 PM, Kristian Høgsberg wrote:
>>>
>>> On Sat, Jan 22, 2011 at 11:42 AM, Benjamin Franzke
>>> <benjaminfranzke at googlemail.com>  wrote:
>>>>
>>>> 2011/1/18 Brian Paul<brian.e.paul at gmail.com>:
>>>>>
>>>>> Before we commit this stuff can you develop a piglit test to test it?
>>>>>
>>>>
>>>> The problem for me is that piglit is GL-only atm. so its not trival to
>>>> write a simple gles2 test-program, as changes to piglit-util would be
>>>> needed.
>>>> It would be important for people wanting to contribute to wayland
>>>> (without having an intel driver, which already advertises
>>>> texture_format_bgra8888) that this is added to mesa without having a
>>>> piglit test yet.
>>>>
>>>> I attach the 3 patches (that were already sent in a bit different
>>>> form) in correct order + 1 Patch that ensures internalFormat==format
>>>> in gles' glTexImage2D.
>>>
>>> I committed the patches for now.  We can add the test when the GLES2
>>> test framework lands.
>>
>> Is there any reason why the extension can't be enabled for regular GL?
>> That might make testing a little easier.
>
> GL_BGRA is already allowed as a format for regular GL glTexImage2D.
> The extension adds it for GLES2 as well, however it for some reason
> requires that the internalFormat must also be GL_BGRA_EXT.  So I'm not
> sure it makes sense to enable for full GL.

OpenGL ES tries to spec glTexImage2D() to minimize the number of image
conversions that the driver needs to implement.  Matching the
internalFormat to the user format should help to minimize the
permutations.

It sounds like the impass to writing a piglit test for this extension
is the ES vs. full GL framework stuff.  If we enable the extension in
full GL we could sidestep that.

-Brian