FreeframeGL and CUDA

Thank you for the details!
graphedit works fine now. However, no matter what Video renderer I connect, I get the same tearing effects. I tried Video Mixing Renderer 9 (7 was not listed) and the other two Video Renderer DirectShow filters. What are the overlay mixers and how to use them in graphedit?

hmm…

so you also get the tearing in graphedit?

“Video Mixing Renderer 7” is one of the two “Video Renderer” that are listed. the other is the old and still default renderer. i’ve never worked with overlay renderers. i think you actually just have to use an overlay-mixer filter before a render…

by the way you are not on vista, are you?

Yes, I get the tearing in graphedit, too.
And no, I’m not on Vista. I run Windows XP.
By the way: I’ve written a small OpenCV test application (without freeframe and VVVV) and when I use the same CUDA code with it as in my VVVV/freeframe project, I don’t get the tearing. So at least my CUDA code seems to be alright.

hmhm…

  • Yes, I get the tearing in graphedit, too.
    that at least sounds like the problem comes not from simultaneous accessing of the GPU from two threads (as we assumend above).

your small openCV test application is completely without directshow i guess? still i have no clue what could cause the tearing then…

  • “that at least sounds like the problem comes not from simultaneous accessing of the GPU from two threads (as we assumend above).”

Why? graphedit uses the GPU, too, doesn’t it?

My OpenCV app doesn’t use DirectShow.
May the tearing have anything to do with render contexts? I’m on the verge of throwing the towel in (concerning to combine CUDA and VVVV)… :(

  • Why? graphedit uses the GPU, too, doesn’t it?
    mmh…maybe…but i thought maybe not with all renderers…or at least it uses the GPU at other times, only when the image is really to be presented. while vvvv is using the GPU seperately in its 3d-rendering-mainloop graphedit doesn’t have this one. but right…still we don’t know when CUDA is using the GPU…so maybe ask in a CUDA forum about the issue?

“so maybe ask in a CUDA forum about the issue?”

I did already, but the only answer I got was that my CUDA code looks alright and works fine without tearing in a small test app of another developer (this test app was not based on VVVV/freeframe).
But thank you for your time and suggestions. Maybe I get an enlightenment about the issue somewhen…

well, good lukc then…