Thanks!
What kind of encoder do you use?
It’s a proprietary H264-complaint encoder implemented using CUDA
You can refer to this document to find one suitable for your use case.
I have reviewed this before, however, as someone who’s never dealt with CUDA/GPU programming, this left me with the question, specifically: if the the memory is unified and allocated on the same physical chip, like on Xavier, why pageable host memory is not accessible on GPU and vice versa, why device memory is not accessible on CPU? I’m not getting this… what “unified” is about this memory then? what exactly does cudaMallocManaged invoke for making its’ allocations seamlessly accessible for both (however, this succinct answer starts shedding some light on this for me)?
I keep reading about this though (btw would appreciate any pointers to good CUDA programming resources, paid or free)…
This might not answer your question but Jetson has a hardware encoder.
Maybe you can check if the following sample can meet your requirements first.
thanks, yes, i’m aware of that one, but the focus for us is currently is on making the implementation we have work on Xavier.