GL_GENERATE_MIPMAP_SGIS and GL_LUMINANCE

Hi,

I have some problems to make the GL_SGIS_generate_mipmap extension work with a one channel texture (GL_LUMINANCE) on a Radeon 9700 pro. The code below produces some black part in the low resolution textures (the 0 level one is the good one) :

glTexImage2D(GL_TEXTURE_2D,0,1,SizeU,SizeV,0,GL_LUMINANCE,GL_FLOAT,First);

If I replace the internal format (1/GL_LUMINANCE8) by GL_RGB8 the mipmaps are generated nicely.

glTexImage2D(GL_TEXTURE_2D,0,GL_RGB8,SizeU,SizeV,0,GL_LUMINANCE,GL_FLOAT,First);

Is there any limitation in the internal formats that are supported by the extension?

Thanks !

[This message has been edited by kard (edited 05-22-2003).]

[This message has been edited by kard (edited 05-22-2003).]

Yes, there are limitations. Read more about it in the spec .

ATI also has a driver bug related to this extension that I found. It’s probably fixed in Cat 3.4, but I haven’t checked:

http://www.opengl.org/discussion_boards/ubb/Forum3/HTML/009011.html

Coriolis: I read the spec, and did not find any limitation concerning the internal format of the base texture. Could you be more precise?

Zeno: It seems to be the same pb… And I test it with the Catalyst 3.4… In your previous topic you mention something about textures with an internal/external format GL_ALPHA when uploaded as GL_FLOAT… Have you try to upload such a texture as non-float? I made some experiment and see that for some strange reasons non square dimensions annihilate the bug…

Thanks !

Cyril-

I’m using GL_LUMINANCE to render my fonts. My standard texture creation path uses SGIS. I have not seen these problems.

I am using Cat3.4. I had a similar problem a few weeks back when getting the new PC, I thought it was ‘my’ code as I edited that section recently. I later switched to ‘Omega’ drivers(for other stuttering problems) and it seemed to fix the problem on my 9700 pro.