videoconvert's performance is too low

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

Hi folks,
I start two gstreamer pipelines using gst-launch-1.0. The difference of the two pipeline is that the second one did color conversion from NV12 to RGBx, as a result, the cpu usage of the pipeline increased from 10% to 94%.

What might the possible cause please? how should I analyze it?

Thanks.

Neil

gst-launch-1.0 -e rtspsrc location="rtsp://172.16.11.155:8554/test" latency=200 short-header=true ! rtph264depay ! h264parse ! omxh264dec ! fakesink

gst-launch-1.0 -e rtspsrc location="rtsp://172.16.11.155:8554/test" latency=200 short-header=true ! rtph264depay ! h264parse ! omxh264dec ! videoconvert ! video/x-raw,format=RGBx ! fakesink

Tasks:   2 total,   0 running,   2 sleeping,   0 stopped,   0 zombie
%Cpu(s): 22.2 us,  4.8 sy,  0.0 ni, 72.5 id,  0.0 wa,  0.0 hi,  0.5 si,  0.0 st
KiB Mem :  8034572 total,  3874304 free,  1273616 used,  2886652 buff/cache
KiB Swap:        0 total,        0 free,        0 used.  6634984 avail Mem

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND                                                                                                                   
 7235 nvidia    20   0 1108596  45440  14876 S  94.0  0.6   1:35.42 gst-launch-1.0                                                                                                            
 7217 nvidia    20   0 1108460  37624  14448 S  10.0  0.5   0:11.39 gst-launch-1.0
2

Yes, videoconvert is CPU only and may use a lot of resources. If power consumption is not a concern, you would boost your jetson with:

sudo nvpmodel -m0
sudo /home/nvidia/jetson_clocks.sh

For converting to BGRx, you may use HW accelerated nvvidconv instead:

gst-launch-1.0 -e rtspsrc location="rtsp://172.16.11.155:8554/test" latency=200 short-header=true ! rtph264depay ! h264parse ! omxh264dec ! <b>nvvidconv</b> ! video/x-raw,format=BGRx ! fakesink
3

Hi Honey_Patouceul:
HW accelerated nvvidconv can decrease the CPU usage. Though videoconvert is CPU only and may use a lot of resources, 84% cpu usage really exceeded my expectation.