PDA

View Full Version : Radeon 9700 texture bug



Zeno
03-12-2003, 01:48 PM
Hi all. I believe I found a pretty serious Radeon 9700 bug but I wanted to post it here first to make sure it wasn't something I was doing wrong.

I'm in the middle of making a fairly large engine hardware-agnostic. When testing on the 9700, I would get a severe lockup (first the app would freeze, then my cursor and windows would freeze a few seconds later) when looking in certain directions. After much hair-pulling and many reboots, I found that the cause of the problem was calling glDrawElements() when I had a DXTC3 compressed texture of size 2048x512 with mipmaps bound to texture unit 0. Resizing this texture to 1024x256 (still DXTC3 with mipmaps) solved the problem.

At the point I did this test, I had made sure that the engine was falling back to vanilla OpenGL....no vertex programs or pixel shaders were being used. Multitexture was the only enabled extension. The Geforce4 Ti4600 ran just fine on the exact same codepath with the exact same data (2048 texture).

Querying GL_MAX_TEXTURE_SIZE on the 9700 returns a size of 2048, so it seems to me that the texture was within the allowed size. Is that size different for compressed textures or for textures with mipmaps? If so, what do I need to call to find that max texture size?

Lastly, some system info: Windows 2000 Pro, Radeon 9700 with Cat 3.1, Pentium IV 1.7 Ghz, 512 Mb RAM.

Anyway, it would be great if this bug could be fixed, or at least not crash windows http://www.opengl.org/discussion_boards/ubb/frown.gif.

Thanks,
Zeno

dorbie
03-12-2003, 02:15 PM
It's more likely to be fixed if you can send code to reproduce it. What happens if they set up this example and it doesn't crash? The more you do to isolate it and allow them to reproduce it the more certain it is that it will be fixed. You have the data and extra info to optimize this to the minimum crashing case.

[This message has been edited by dorbie (edited 03-12-2003).]

Zeno
03-12-2003, 02:18 PM
Ok. If you can't see anything I'm doing wrong, I'll try to isolate it and submit it tonight.

I think that there's utility in posting the problem here first, though. This way, it'll show up in a search for anyone else who may be having the same problem and you guys can do a first-pass sanity check on the problem before I bother ATI about it.

-- Zeno

[This message has been edited by Zeno (edited 03-12-2003).]

Dodger
03-12-2003, 04:45 PM
Agree, Zeno - posts about potential bugs on here have saved me hours on end of debugging more than once. If you get a reply from ATI, please do post a followup with the resolution.

Zeno
03-12-2003, 06:49 PM
Ok, I've isolated the problem and uploaded an example app here:
http://www.sciencemeetsart.com/wade/temp/ATITexBug.zip

I also emailed devrel@ati.com. I hope that's the appropriate place to file the bug report.

It turns out that mipmaps are unimportant. A 2048x512 2D compressed texture will cause the app to freeze and take windows down with it on a 9700 Pro. Both versions work fine on Geforce3/4.

Anyone who wants to compile and/or run this program, please be aware that it may cause your computer to crash. You have been warned http://www.opengl.org/discussion_boards/ubb/wink.gif

-- Zeno


[This message has been edited by Zeno (edited 03-12-2003).]

Ysaneya
03-13-2003, 12:38 AM
Worked fine on my Radeon 8500.

Y.

abrodersen
03-13-2003, 12:52 AM
I have experienced the same problem on my Radeon 9700, using textures of width or height 2048.

Hopefully this is fixed in the Catalyst 3.2 drivers they claim to have released yesterday (Still trying hard to find out exactly where these drivers were released, as all links on their pages point to the old 3.1 drivers ;-)

Anders

davepermen
03-13-2003, 01:52 AM
Originally posted by abrodersen:
Hopefully this is fixed in the Catalyst 3.2 drivers they claim to have released yesterday

they have released 3.2 catalist. but catalist is not just drivers. drivers stay the same, for now. update will follow shortly, they say. can't you read? its all written on their page, even with a pop up window!!

abrodersen
03-13-2003, 02:40 AM
Originally posted by davepermen:
they have released 3.2 catalist. but catalist is not just drivers. drivers stay the same, for now. update will follow shortly, they say. can't you read? its all written on their page, even with a pop up window!!

I haven't seen that anywhere, and I don't get any popups! Not with Mozilla, not with IE. However, I can find several links to their "unified driver" release 3.2, only they all point to the 3.1 drivers!

Also, from the official press release:


ATIís newest unified driver release CATALYST 3.2 can be found at http://www.ati.com.


You might interpret this differently than I do, but if you ask me, I'd say this should mean they have released a new set of drivers!

Anders

Zeno
03-13-2003, 09:10 AM
I got a response back from ATI developer relations....damn that was fast.

They said that it did indeed cause a system hang on Catalyst 3.1, but that it works fine on Catalyst 3.2.

Thanks to ATI for the fast response http://www.opengl.org/discussion_boards/ubb/smile.gif

-- Zeno