Announcement

Collapse
No announcement yet.

RadeonSI Getting Fixed Up To Expose 10-bit VP9 Decode

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • agd5f
    replied
    Originally posted by perpetually high View Post

    I've read that Polaris/RX480 has hybrid VP9 support, not sure on the Linux support yet.
    Raven (VCN 1.0) is the first chip with VP9 support in hardware. Polaris chips do not have VP9 support in hardware.

    Leave a comment:


  • Guest
    Guest replied
    Originally posted by tildearrow View Post

    Even copying a surface to another surface twice doesn't work. Wow.
    ??? vaCreateSurfaces allocates memory for decoding/encoding, vaDeriveImage merely creates a VAImage object that represents a given surface and doesn't perform any copies.

    Leave a comment:


  • tildearrow
    replied
    Originally posted by sandy8925 View Post
    Lol, VP9 decode? Even H.264 decode through VAAPI doesn't work. AMD's vaCreateSurfaces implementation creates NV12 surface, while their vaDeriveImage implementation doesn't support NV12.
    Even copying a surface to another surface twice doesn't work. Wow.

    Leave a comment:


  • perpetually high
    replied
    Originally posted by sandy8925 View Post

    Does your GPU support VP9 decode?
    I've read that Polaris/RX480 has hybrid VP9 support, not sure on the Linux support yet.

    Leave a comment:


  • aufkrawall
    replied
    Originally posted by sandy8925 View Post
    Lol, VP9 decode? Even H.264 decode through VAAPI doesn't work. AMD's vaCreateSurfaces implementation creates NV12 surface, while their vaDeriveImage implementation doesn't support NV12.
    Except that it works?

    Leave a comment:


  • Guest
    Guest replied
    Originally posted by perpetually high View Post
    Does anyone know if this would solve the following error when using mpv to play a 4K youtube video?

    Code:
    $ mpv 'https://www.youtube.com/watch?v=Bey4XXJAqS8'
    Playing: https://www.youtube.com/watch?v=Bey4XXJAqS8
    (+) Video --vid=1 (*) (vp9 3840x2160 29.970fps)
    (+) Audio --aid=1 --alang=eng (*) 'DASH audio' (opus 2ch 48000Hz) (external)
    Subs --sid=1 --slang=en 'vtt' (webvtt) (external)
    mesa: for the -simplifycfg-sink-common option: may only occur zero or one times!
    mesa: for the -global-isel-abort option: may only occur zero or one times!
    [B][ffmpeg/video] vp9: No support for codec vp9 profile 0.
    [ffmpeg/video] vp9: Failed setup for format vaapi_vld: hwaccel initialisation returned error.[/B]
    AO: [pulse] 48000Hz stereo 2ch float
    VO: [gpu] 3840x2160 yuv420p
    AV: 00:01:28 / 00:29:36 (4%) A-V: 0.000 Cache: 80s+155MB
    Does your GPU support VP9 decode?

    Leave a comment:


  • Guest
    Guest replied
    Lol, VP9 decode? Even H.264 decode through VAAPI doesn't work. AMD's vaCreateSurfaces implementation creates NV12 surface, while their vaDeriveImage implementation doesn't support NV12.

    Leave a comment:


  • bridgman
    replied
    Originally posted by aufkrawall View Post
    Hm, then perhaps someone qualified should make a proposal to VAAPI devs how it could be get to work on AMD?
    Reading through the bug ticket comments it seems that has been done a couple of times already, but will try to confirm offline.

    Leave a comment:


  • aufkrawall
    replied
    Originally posted by bridgman View Post
    With respect, the bug ticket seems to suggest that is a limitation in VA-API, not AMD driver/HW.
    Hm, then perhaps someone qualified should make a proposal to VAAPI devs how it could be get to work on AMD? It works on Intel, just tested it with my Gemini Lake device (btw. VAAPI decoding of VC-1 interlaced is still broken on AMD too, producing garbled output).
    Of course I'm not mad about this issue, as I mainly consume web content, which is entirely progressive.

    What bothers me much more is this problem of the xf86-video-amdgpu driver + wine (yeah, I've linked it before):


    Also reported it to wine in the meantime:


    Completely destroys Wine gaming with FreeSync on modern AMD cards such as Vega or Navi, and is also a bad hassle otherwise.

    Leave a comment:


  • bridgman
    replied
    Originally posted by ihatemichael
    They fix a lot of bugs, although there are some long standing bugs on the bug tracker as well.
    Fixed that for you. The existence of an unfixed bug does not mean that other bugs are not being fixed.

    Leave a comment:

Working...
X