View Full Version : Ati catalyst 5.6 NPOT texture bug

06-17-2005, 08:48 AM

Im using the NPOT function of opengl 2.0, that is only call glteximage2d as normal but with non power of 2 values for width and height.

I get really strange bugs in the textures after loading several small images ( 120*160px) and one bigger. Guess many different combinations might work the same here, but on my bigger images, 1536*1024, they doesnt appear.

The bug looks like this (bug) (http://www.annat.nu/images/oglbug.jpg)
and it seems to be some problem with overwrites in the memory (just load one texture works fine, but with more textures i get more of these problems )

A know bug, or can it be me?

06-17-2005, 01:33 PM
I'm not aware of any such bugs, do you have a repro-case?

06-18-2005, 12:34 AM
file here (http://www.annat.nu/files/ati_strange.zip)

Unzip and start, and make the window repaint itself once more( rezize or hide/activate)

All source should be there, and i hope you know delphi :)

if i change the texture upload to

powWidth:=trunc(intPower(2,Ceil(Log2(bmp.Width)))) ;
powHeight:=trunc(intPower(2,Ceil(Log2(bmp.Height)) ));
glTexImage2D(GL_TEXTURE_2D ,0,GL_RGB8, powWidth, powHeight, 0, GL_RGBA, GL_UNSIGNED_BYTE, nil);
glTexSubImage2D(GL_TEXTURE_2D,0,0,0,bmp.Width, bmp.Height,gl_RGBA,GL_UNSIGNED_BYTE,bmp.ScanLine[bmp.height-1]);
instead of using just NPOT then i dont get the same artifacts.. but if i make my window very small (60*8 pixels or so of clientarea) the drawn image becomes very disturbed again.

06-18-2005, 12:53 AM
I guess you want the specification as well.

I have a 9800pro. a very old one, the first i could get my hands on here in sweden.

06-18-2005, 02:25 AM
I got similar disrupted textures (mainly cube maps) on X300 (PCI-Express) and 9600 mobility (notebook) systems (at least for driver versions 5.1 to 5.5). Will try to create some snapshots on monday at work, since I have no small repro-case handy yet...

06-20-2005, 07:25 AM
Thanks for the repro Mazy, I'll get this bug reported.

06-20-2005, 08:08 AM
Here are 2 screenshots of this bug in action:

picture #1 (http://cvscgs.hpi.uni-potsdam.de/baumann/pics/bad-cubemap-01.jpg)
picture #2 (http://cvscgs.hpi.uni-potsdam.de/baumann/pics/bad-cubemap-02.jpg)

The disruption often happens after enlarging the viewport (e.g. during maximizing the window); you also can see that some of the VBO vertex arrays are lost at the same time... :-(

06-20-2005, 11:44 AM
Are you using small NPOT texture as well?

06-21-2005, 12:07 AM
The cubemaps are certainly not NPOTs ;) , but the aerial images draped over the terrain are NPOTs (or to be more precise: GL_EXT_texture_rectangle textures).

06-22-2005, 05:40 AM
FYI, this bug has been fixed.

06-22-2005, 06:50 AM
Im suffering from a similar problem. In my case I create a number of pbuffers (3 different contextes get shared among groups of them). The first time the pbuffers get created, everything is ok. Then I delete all of them and re-create the whole thing again. During this, no GL Errors or other wgl-related problems occur. Unfortunately, then I get problems looking like the ones depicted in this thread. No GL-errors occur, everything seems to work fine, except it looks horrible.

Might this bug be related to the npot bug? When is the fixed driver available?

06-22-2005, 08:19 AM
It could be the related, but I wouldn't be too sure. If you have a repro-case I'd like to look at it to see if it's been fixed.
From fix to public release you can count on a few months.

10-25-2005, 05:13 AM
@Humus: the error still exists in the Catalyst 5.10 driver. Can you check, if the fix should be contained in the 5.10 driver or do we have to wait for the 5.11? Any release date yet?

10-26-2005, 04:22 AM
This bug was fixed in Cat 5.9. If you're still seeing a problem it's another bug. Could you please send me a repro case to epersson 'at' ati.com?