Zink Seeing VA-API Video Acceleration Implemented Over Vulkan Video

[phoronix]

Phoronix: Zink Seeing VA-API Video Acceleration Implemented Over Vulkan Video

The Zink OpenGL-on-Vulkan driver has experimental code now available for testing that also implements the Video Acceleration API (VA-API) atop the Vulkan Video APIs. This is an interesting effort that now allows VA-API applications to rely on drivers with Vulkan Video support underneath...

https://www.phoronix.com/news/Zink-VA-API-On-Vulkan-Video

[Beryesa]

This might be the long-term solution for Nvidia video acceleration on Firefox/Chromium? Given they'll probably take a decade to support Vulkan video themselves ╮⁠(⁠＾⁠▽⁠＾⁠)⁠╭

[ahrs]

This might be the long-term solution for Nvidia video acceleration on Firefox/Chromium? Given they'll probably take a decade to support Vulkan video themselves ╮⁠(⁠＾⁠▽⁠＾⁠)⁠╭

Zink already works with Firefox, I've tested that before but you don't get video decoding because it relies on VAAPI. If this actually works then it'll be very nice for things like the Apple Silicon graphics driver that got merged recently (can't remember its name, it's not important) and Nouveau, both of which have very poor OpenGL support and are focusing all of their efforts on Vulkan.

I don't know why me as an AMDGPU user would use this when they have an excellent OpenGL driver though.

[dlq84]

Zink already works with Firefox, I've tested that before but you don't get video decoding because it relies on VAAPI. If this actually works then it'll be very nice for thinks like the Apple Silicon graphics driver that got merged recently (can't remember its name, it's not important) and Nouveau, both of which have very poor OpenGL support and are focusing all of their efforts on Vulkan.

I don't know why me as an AMDGPU user would use this when they have an excellent OpenGL driver though.

He's talking about video decode acceleration, not graphics.

[patrick1946]

How does that work with energy-saving measures? AFAIK, the video part of the chip is very often independent of the 3D part. So, you can watch a full-screen video without enabling the 3D part of a GPU. Can Vulkan work in that way?

[dragon321]

This might be the long-term solution for Nvidia video acceleration on Firefox/Chromium? Given they'll probably take a decade to support Vulkan video themselves ╮⁠(⁠＾⁠▽⁠＾⁠)⁠╭

NVIDIA already supports Vulkan Video. Hardware video decode also works on Firefox on NVIDIA with nvidia-vaapi-driver that implemented VAAPI on top of NVDEC.

[Britoid]

This is a cool idea but I imagine it will always be more power efficient to use dedicated APIs as they can use on-device decoders.

[davidbepo]

This is a cool idea but I imagine it will always be more power efficient to use dedicated APIs as they can use on-device decoders.

vulkan video uses the HW blocks, so this does too

[Beryesa]

NVIDIA already supports Vulkan Video. Hardware video decode also works on Firefox on NVIDIA with nvidia-vaapi-driver that implemented VAAPI on top of NVDEC.

Yes, nvidia has vulkan video, I meant Firefox adding vk video will take forever.
Vaapi drivers for Nvidia aren't that good afaik (I have hybrid, I use amdgpu for browsers anyway) I mean as a long term solution etc.

[ahrs]

He's talking about video decode acceleration, not graphics.

I know, right now the status quo is that you get graphics acceleration with Zink but not video acceleration. Implementing VAAPI on top of Vulkan Video should hopefully fix that (in the future at least, I don't know if Nvk and the Apple Silicon Vulkan drivers support Vulkan Video yet, but they surely will in the future).