I need to use gst clockoverlay in my application to timestamp files while recording. But the problem is that the clockoverlay element causes the CPU usage to go from 25% while recording fullHD@60fps to 115% with timestamping!! Obviously, this is unacceptable because I need to record multiple channels simultaneously and perform a bunch of other tasks.
I constructed a simple passthrough pipeline below. This has a CPU usage of 72% while it's just 2% without the overlay.
Is there a way to do hardware composition so that the CPU doesn't get so heavily involved. Can the overlay be mapped to the MALI GPU? Any other suggestions are welcome. The only restriction is that I cannot overlay the timestamp in HW on the input path because the timestamping is only for recorded files and the live display should show the original video.