For input buffers there are two ways of allocating: NvEncCreateInputBuffer or NVEncRegisterResource.
But for output buffers there is only NvCreateBitstreamBuffer.
I’m trying to avoid memcopy on host, I already have shared, pinned and cuda-registred host memory for output of various codecs, I would like to use it with nvenc too.
Other way would be to have the nvenc output to device memory, then I can copy it myself to host.
Is something like that possible?