I want to combine your example of ARGUS CudaBayerDemosaic with some nppi processing. But everytime I call nppi function my stream crashes. (while the nppi function run until end and produce nice results)
I am calling the nppi functions from another cpu thread and I not settings streams until now, so everything should be synchronized in the default stream.
I guess it has to be something about the cuda context.
In your example there is an explicit creation of a cuda context. So in my newly spawned thread I use cuCtxSetCurrent to bind that context to my thread. This way normal CUDA kernels work.
But when using NPPI it crashes. But I did not find a way to set cuda context for nppi. So I thought maybe it just uses the primary context and the primary context fails with my newly created one.
So I tried to use primary context with the egl stream, but that way I get (CUresult invalid device context).
So what would be the correct way to bring this example together with nppi?
Thanks for your help!