[Mesa-dev] _mesa_texstore_argb8888 with GL_RGBA -> GL_RGB

[Kenneth Graunke]

On 04/25/2011 04:36 PM, Tristan Schmelcher wrote:
> On 25 April 2011 16:22, Ian Romanick<idr at freedesktop.org>  wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On 04/25/2011 04:15 PM, Tristan Schmelcher wrote:
>>
>>> I have an OpenGL app where I'm writing video frames to textures. The
>>> video frames are semantically opaque but the alpha in the source
>>> pixels is set to 0xff, so I have been using glTexImage2D with
>>> format=GL_RGBA and internalFormat=GL_RGB to indicate that the input
>>> has a valid alpha channel but that the OpenGL implementation could
>>> discard it if it wanted to (e.g., if the hardware had native support
>>> for GL_RGB which was faster than its GL_RGBA support, or something).
>>
>> I can't imagine that ever happening for 8-bit per component color.  RGB
>> with 8-bits per component is 3 bytes, so drivers will always promote
>> this to RGBx (a single byte of unused padding).  This is basically what
>> you already have.
>
> Yeah, I am definitely going to change my code to use GL_RGBA to fix
> things ... I'm just puzzled about why the perf is so bad on this
> system when using GL_RGB. My app has been used on a lot of other
> systems (with many different OpenGL implementations) and this is the
> first time I've seen this problem.
>
>>
>>> This has worked well for me on most systems, but today I found that it
>>> results in poor glTexSubImage2D() performance on an Intel IGDNG_M GPU
>>> (i965-based) with MESA 7.7.1 (using Ubuntu Lucid). Because
>>> internalFormat is GL_RGB instead of GL_RGBA, _mesa_texstore_argb8888
>>> takes a slow path by using _mesa_swizzle_ubyte_image instead of
>>> memcpy_texture.

If you have a debug build of Mesa, you could try running your app with 
INTEL_DEBUG=fall set.  That should print out warnings if you hit any 
software fallbacks in our driver.  They should be very rare, but would 
cause a performance cliff.

I know that I fixed a software fallback when doing glCopyTexImage from 
ARGB8888 to XRGB8888 in master last month.  Not sure if it's relevant.

As Ian mentions, a lot has changed since 7.7.  Even if that one function 
hasn't changed much, it would still be worth a try.

--Kenneth