Hey, I’m trying to use PBOs to accelerate texture uploads. This works when I first fill my own local buffer, and then call BufferData to copy it to host memory, but when I use Map, and then try to write the data straight into the host memory, then nothing happens (ie. the contents of the memory remain unchanged).
Also, I’ve just looked at the GPU particles demo in the nVidia SDK, which is supposed to use PBOs for reading the texture into the vertex buffer, but it doesn’t work on my GF6600(77.72)… I cannot see any particles. I’m guessing that the two problems could be related.
Could you guys test the particles demo and see if it works? Maybe it’s just my setup that’s behaving like this?
Could someone try the nVidia GPU particles demo, and report if it works on a GF6600, please?
You can find it here !
got the same card + driver and it is NOT working either…
Could someone from nVidia look into this pls?
Hmmm let me talk to our engineers about this.
Yes, there’s a VBO bug with the 77.72 driver and the gpu_particles demo. It will be fixed in the next release.
In the meantime, you can work around the problem by looking for the function “InitNewParticles” in ParticleSystem.cpp and changing the #if so that it uses glBufferData rather than mapping the array.
LOL! Perhaps another reason to not map?
The most obvious reason to not map I think of, is that I consider the server really remote (even if on the same machine). NUMA++.
Hah, thanks for the confirmation! It must be the same problem that I have in my prog then. At least now I can laid back a bit and not go crazy trying to debug my code, thx!
Yeah it didn’t work on 77.77 either.
Am i the only one who is scared by that kind of thing ? If NVidia releases a new driver, i’d expect their QA & Testing department to at least test all the demos and samples they have officially released on their own website ? How long does it take to have one engineer loop through all the demos, simply run them, and make sure everything still works ? ATI is no better in that way either. That’s quite incredible.
I think it’s all automated these days - there’s banks of PC emulators, each with a different hardware/software configuration, all running a fixed batch of test programs. The resulting framebuffer is probably compared with a reference framebuffer, and differences are flagged.
These test programs probably won’t include the demos off their website.