I was looking for this mess of hw encoders on ARM singleborad computers and randomly hit this blog: https://www.willusher.io/general/2020/11/15/hw-accel-encoding-rpi4 so apparently default 'OS' on Pi4 uses v4ĺ2 encoder, should be easy to create profile for (just copy vaapi one and edit ) and as bonus you can find python + ffmpeg encoding manager .... another random project is wayland capture program, might not work on intel due to fb modifiers (?) https://trac.ffmpeg.org/ticket/8542 https://github.com/cyanreg/txproto and all this because adventurious user 'goat' reported he had better luck with cin-gg on arm single-board computer under Wayland/Xwayland - namely timeline scrolling performance.
and all this because adventurious user 'goat' reported he had better luck with cin-gg on arm single-board computer under Wayland/Xwayland - namely timeline scrolling performance. --
Wow, that is interesting and I really wonder why that would be -- it must be some implementation feature of Wayland ?
пт, 21 окт. 2022 г., 21:11 Phyllis Smith <[email protected]>:
and all this because adventurious user 'goat' reported he had better luck with cin-gg on arm single-board computer under Wayland/Xwayland - namely timeline scrolling performance. --
Wow, that is interesting and I really wonder why that would be -- it must be some implementation feature of Wayland ?
I think yes? On most modern systems X server uses 3d acceleration via glamour module, and 3d part on those singleboard computers tend to use even more weird layout for their buffers - so you can't just read texture via cpu (memcpy), it must be detiled, sometimes in software ... and wayland does some of this via dma-bufs (?) kernel level shared buffer implementation .. so it can do things not defined or allowed in Opengl/egl or so I gathered from forums ....
пт, 21 окт. 2022 г., 21:16 Andrew Randrianasulu <[email protected]>:
пт, 21 окт. 2022 г., 21:11 Phyllis Smith <[email protected]>:
and all this because adventurious user 'goat' reported he had better luck with cin-gg on arm single-board computer under Wayland/Xwayland - namely timeline scrolling performance. --
Wow, that is interesting and I really wonder why that would be -- it must be some implementation feature of Wayland ?
little video demo (hopefully yandex links actually watchable with their integrated player) https://www.linux.org.ru/forum/linux-hardware/16959253/page3#comments (for some reason IceWM recording appear more smooth than it was in reality)
I think yes? On most modern systems X server uses 3d acceleration via glamour module, and 3d part on those singleboard computers tend to use even more weird layout for their buffers - so you can't just read texture via cpu (memcpy), it must be detiled, sometimes in software ... and wayland does some of this via dma-bufs (?) kernel level shared buffer implementation .. so it can do things not defined or allowed in Opengl/egl
or so I gathered from forums ....
participants (2)
-
Andrew Randrianasulu -
Phyllis Smith