I'm increasingly excited about what #opencl support coming together across the board in #mesa could mean for multimedia in the free ecosystem.
Just some thoughts about what I *think* is on the table:
- fluent/efficient AV1[1] playback across the board, *maybe* even on hardware like the Raspberry Pi4
- real-time AV1 encoding in reasonable quality for video chat, screencasting and camera recording
- a GPU ISP, allowing reasonable image quality with any camera sensor
1: or any other format?
This is probably over-optimistic - but seeing what's being done with CUDA on that other platform, how dmabuf and kms are evolving and how gl/vk support came together in Mesa - it might be somewhat realistic?
Hm, looks like I was wrong about de- and encoding parts and it's actually hard to speed those up with opencl. One claim I read was that modern codecs are just too branchy and CPU ends up being better - does anyone know more about that, specifically regarding AV1?