i'm running on gst 1.8.3. I'm using it for displaying multiple videos
concurrently in combination with OpenGL. So the basic flow is: I open
pipelines, grab some frames with an appsink and push these to OpenGL. More
or less comparable to a simple pipeline flowing into an glimagesink but i'm
using some postprocessing on the GL side and of course to convert different
pixel types on the GPU.
I was running some performance tests with the bunny video as my application
fails to play this one smoothly. My app runs at about 70% on the cpu playing
this single video. gst-launch with a simple pipeline like filesrc !
deocdebin ! glimagesink takes about the same effort on the cpu.
If one now changes glimagesink to autovideosink on the above pipeline
gst-launch just takes 10% on the cpu. I took a look at the pipeline plot and
noticed that autovideosink wraps a vaapisink in contrast to appsink and
glimagesink which gives it a huge performance boost.
For me it would make big sense to make the appsink use the vaapi benefits,
if available, as it's the most versatile. Are there any plans to integrate