NVIDIA GPU Direct DVP library dvpInitCUDAContext return DVP_STATUS_UNSUPPORTED or DVP_STATUS_INVALID_DEVICE

Accelerated Computing GPU-Accelerated Libraries
1

Hello,

I’m a student new to CUDA, new to GPU, and new to BlackMagic.
I would like to use GPU Direct to send the frame on the GPU as soon as I receive it.
I tried to copy from the LoopThroughWithOpenGLCompositing but this sample doesn’t copy using CUDA but instead OpenGL.
I tried to use dvpInitCUDAContext from the DVP library, but I can’t find any sample that use it.
dvpInitCUDAContext is returning DVP_STATUS_INVALID_DEVICE so I use cudaSetDevice(0) to initialize CUDA, and then I get DVP_STATUS_UNSUPPORTED.

I don’t have any idea how to do it using CUDA.

2

This isn’t the right forum for support on BlackMagic. That is a blackmagic sample code. Likewise, DVP is also a blackmagic technology. You may wish to ask on blackmagic forums.

3

Hello,

Thank you.

4

Hello Robert,

I think there’s a misunderstanding here…

According to the DeckLink manual…

NVIDIA GPUDirect support
NVIDIA GPUDirect is supported on Windows and Linux for x86_64 architecture where those platforms are
also supported by NVIDIA. GPUDirect support requires the use of the DVP library supplied by NVIDIA.
See the LoopThroughWithOpenGLCompositing for a detailed example of integrating the DeckLink API
and NVIDIA GPUDirect

BMD Decklink manual - Page 15 - 1.1.3.1

So, full disclosure… NVIDIA does gatekeep this functionality… When you email requesting access to it… never get a response… either with personal email or with business email…

5

Correct, sorry for the mixup. DVP is NVIDIA. The sample code is from blackmagic. Regarding the rest,

See here. It doesn’t look like email is the right way to request access, and the NVIDIA DVP developers appear to only intend to give direct access to 3rd party hardware manufacturers, not individual developers. Individual developers should work with the 3rd party and their SDKs.