0

I got a orangepi5 that has hardware decoders for h264,hevc and other formats. I am using armibian Armbian 23.5.5 Jammy CLI kernel legacy 5.10.160. I've installed a custom ffmpeg that uses RockChip MPP and hence enables those decoders.

I am trying to use those hardware decoders to create a loopback device at /dev/video0. For example, using the following command:

ffmpeg -rtsp_transport tcp -y -c:v  h264_rkmpp_decoder -re -i rtsp://admin:passwd@ipcam-front-up:554/stream=0 -c:v rawvideo -pix_fmt yuv420p -f v4l2  /dev/video0 

It works but if I look at htop I see 12 processes/threads besides the main thread/process that spawned them (by their process group number) and I get ~7% of average CPU usage.

If I am not transcoding anything, why that? If I for example use only the bellow I get only 1% of average CPU usage .

ffmpeg -rtsp_transport tcp -y -c:v  h264_rkmpp_decoder -i rtsp://admin:passwd@ipcam-front-up:554/stream=0 -f null -

Why -f v4l2 is creating that much overhead. Or what am I doing wrong? I just want to make that stream available for another software (motion project) that cannot use that hardware decoder directly.

0

You must log in to answer this question.

Browse other questions tagged .