I’m writing an occlusion culling program and there’s a weird problem with my ATI driver under Fedora 2. I have a Dell Inspiron 8600 with an ATI Radeon mobility 9600 card on it.
I installed the newest ATI driver (3.14.6) on my system - it works correct, well but not the best performance, only 1500fps in glxgears…but when I try to enable the occlusion culling extensions in my code, the program crashes. If I compile the same code at windows XP OS, everything works fine.
please contact me if you know more about this problem - I believe it’s a driver bug, but i’m not sure.
I haven’t heard any problems with occlusion culling in ATI’s drivers, but that doesn’t mean that there aren’t any. Which occlusion test extension are you using? Therearethree .
Can you be more specific about how the program crashes? Is it in your code? Is it in driver code? Can you provide a back trace? If you have a small glut or SDL based program that exhibits the problem, that would also help.
Finally, what does the app do with Mesa’s software rendering (other than run slowly )?
thhank you for your reply.
you’re right - there are three. I use the ARB_OCCLUSION_QUERY extension.
I’m using the extenson in a QT software, but there is one small (GLUT) programm from Brian Paul I tried before and it didn’t work either.
I can’t say exactly where the program crashes, but I can send the queries, but when I wait for the result (glGetQueryObjectivARB(OccQuery, GL_QUERY_RESULT_AVAILABLE_ARB, &ready) I can wait until I die…
I don’t have any problems with the code when I’m using NVidia cards or Windows OS.
As far as I can remember the occlusion query demo in the nvidia sdk didn’t work correctly either with the ati drivers (but it did not crash).
A few driver versions seemed to return just a random count for the number of rendered fragments. I think the 3.14.6 (can’t test right now) always returns -1.