slow decoding performance

I’m currently using the following pipeline to decode 1080p video feed from my web camera.

/usr/bin/gst-launch-1.0 -vvv \
  udpsrc address=192.168.0.94 port=5005 ! \
  video/x-h264,width=1920,height=1080,framerate=30/1,stream-format=byte-stream ! \
  queue ! \
  h264parse ! \
  omxh264dec ! \
  nveglglessink

I’m using the above as a proxy to test the final pipeline in my application which is:

udpsrc address=192.168.0.94 port=5005 ! \
  video/x-h264,width=1920,height=1080,framerate=30/1,stream-format=byte-stream ! \
  queue ! \
  h264parse ! \
  omxh264dec ! \
  nvvidconv ! \
  appsink

The source comes from another pipeline which is as follows:

/usr/local/bin/gst-launch-1.0 -v \
  uvch264src \
    iframe-period=100 \
    initial-bitrate=24000000 \
    average-bitrate=24000000 \
    device=/dev/video0 \
    auto-start=true \
    name=src src.vidsrc ! \
  video/x-h264,width=1920,height=1080,framerate=30/1,profile=constrained-baseline ! \
  h264parse ! \
  rndbuffersize min=1500 max=1500 ! \
  multiudpsink clients=192.168.0.162:5004,192.168.0.94:5005

On my desktop I can decode the source at full frame rate without any issues, but on the TK1 it only renders at approx 5 fps.

Any idea what I could do to improve the performance of decoding?

I use the following to max perf on my tk1:

echo -1 > /sys/module/usbcore/parameters/autosuspend
echo 1 > /sys/kernel/debug/clock/override.gbus/state
echo 852000000 > /sys/kernel/debug/clock/override.gbus/rate
echo "gpu set"
echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
echo "scaling set to performance"
echo 0 > /sys/devices/system/cpu/cpuquiet/tegra_cpuquiet/enable
echo "disabled quiet mode"
echo 1 > /sys/devices/system/cpu/cpu0/online || true
echo "enabled cpu 0"
echo 1 > /sys/devices/system/cpu/cpu1/online || true
echo "enabled cpu 1"
echo 1 > /sys/devices/system/cpu/cpu2/online || true
echo "enabled cpu 2"
echo 1 > /sys/devices/system/cpu/cpu3/online || true
echo "enabled cpu 3"

Hi,
Please try ‘nveglglessink sync=false

Thanks that helped a lot, hopefully this framerate will transfer over to using an appsink with the same parameter.