How does the data above DMA carry out zero-copy encoding?

I refer to the coding example in the tegra_multimedia_api. I want to directly map to the encoded capture_plane via DMA. How do I use the zero-copy encoding?
Is there any way to get video data in DMA into the encoder without incurring extra overhead?

I want to convert video data in DMA to EGLStream data and then import it into gstreamer. Is there any way to do this at low cost?


Please check tegra_multimedia_api\samples\v4l2cuda in multimedia API first.
It demonstrates v4l2 source -> CUDA buffers, which should be helpful for your use case.