Thanks Microsoft: RadeonSI Enables Async VA-API For Faster Video Acceleration

[phoronix]

Phoronix: Thanks Microsoft: RadeonSI Enables Async VA-API For Faster Video Acceleration

Last week I wrote about how Microsoft landed a VA-API improvement in Mesa to support faster Video Acceleration API encoding with FFmpeg. That code was initially only wired up for the Microsoft D3D12 driver within Mesa for WSL use-cases, but now AMD has taken advantage of the new capability for RadeonSI Gallium3D usage with their Radeon GPUs...

https://www.phoronix.com/news/RadeonSI-VCN-Async-Mesa-22.3

[Linuxxx]

That's nice and all, but still doesn't change the fact that I'm unable to activate hardware decoding for mpv's "gpu-next" video output on my AMD Radeon R9 380, whereas doing the same on my Intel i3-7100U works flawlessly.

Other AMD users on here already reported the same problem, so at least I know I'm not alone with this issue, which is strangely reassuring...

[fagnerln]

Weird times 🤣

[mick3]

Someone already asked this at the MR and the committer said async VA-API might also work with the previous Video Code Engine VCE unit which all AMD CGN graphics cards up to Vega10/20 use for H.264/H.265 encoding.

AMD just won't be able to verify this.

Are there any other Mesa RadeonSI developers around who could help AMD to verify this with their pre-VCN graphics cards?

Thanks,
M.

> the AMD change is just two lines of code for enabling the video capability with Raven graphics hardware and newer.
> Basically, AMD Radeon graphics with the Video Core Next "VCN" block can make use of this async functionality.

[IActuallyKnowItAll]

Still doesn't change the fact that MS is evil.

[Toggleton]

That's nice and all, but still doesn't change the fact that I'm unable to activate hardware decoding for mpv's "gpu-next" video output on my AMD Radeon R9 380, whereas doing the same on my Intel i3-7100U works flawlessly.

Other AMD users on here already reported the same problem, so at least I know I'm not alone with this issue, which is strangely reassuring...

With a rx550 i have no problem with mpv with gpu-next. i see for h264/h.265 vaapi-copy. so maybe it is specific to that older gen? mesa, mpv and ffmpeg all compiled from git on Arch. I think the precompiled mesa version on Arch did not really work with vaapi for me.

[tildearrow]

Still doesn't change the fact that MS is evil.

Why does Phoronix thank the evil? Sure, sometimes Microsoft and Google make useful contributions, but yeah.

[Volta]

Why does Phoronix thank the evil? Sure, sometimes Microsoft and Google make useful contributions, but yeah.

Maybe 'Phoronix' thought he'll make 100+ comments here. It seems nobody cares. What a pity.

[marlock]

A Star-Treck Borg version of Linus Torvalds looks at Microsoft and says: "resistance is futile, you will be assimilated"

The EEE article on Wikipedia will now need a disambiguation: "Embrace Extend and Extinguish" for the classic MS era, then "Embrace, Extend and be thanked and Extinguished" after Linux reached critical mass and became too big for MS to keep up with

IMHO the subtext to Michael's post is ironic towards Microsoft's attempt to extend Mesa inly towards Windows WSL and inadvertedly helping proper Linux get better, but some folks seem to have missed the joke.

There is no need for him to explain his entire position on the matter at each and every post... you can have a sense of his positions from previous remarks on WSL and stuff geared/developed for windows first or exclusively. Those are technologically interesting efforts but clearly show Microsoft's onesidedness, and the articles are all careful to note the absence of usage outside WSL for proper Linux when that happens.

On the other hand this is so intrinsicly funny that the thank-you could be considered overkill, lol

[Linuxxx]

With a rx550 i have no problem with mpv with gpu-next. i see for h264/h.265 vaapi-copy. so maybe it is specific to that older gen? mesa, mpv and ffmpeg all compiled from git on Arch. I think the precompiled mesa version on Arch did not really work with vaapi for me.

VAAPI-copy isn't the problem, however plain "hwdec=vaapi" is.

Could you please try the zero-copy path and see whether that one works on your RX550 aswell?

Thanks!