https://bugs.freedesktop.org/show_bug.cgi?id=68792

          Priority: medium
            Bug ID: 68792
          Assignee: dri-devel@lists.freedesktop.org
           Summary: Problems during playback of h264 files using UVD and
                    VLC on AMD E-350 CPU
          Severity: normal
    Classification: Unclassified
                OS: Linux (All)
          Reporter: rvdb@cryptolab.net
          Hardware: x86-64 (AMD64)
            Status: NEW
           Version: git
         Component: Drivers/Gallium/r600
           Product: Mesa

Created attachment 84971
  --> https://bugs.freedesktop.org/attachment.cgi?id=84971&action=edit
vlc -vvv output log

I have a small system with a E350 CPU in it and try to use a recent git pull of
VLC, mesa and xfree86-video-ati, on a 3.11rc6 kernel, to run with VDPAU and the
UVD chip on the CPU.
VLC indeed manages to play H264 files in DVD resolution but :

1/ It seems to fail on color space matrix :

[0x7ffa0f00c288] vdpau_chroma filter debug: using video mixer sharpness feature
[0x7ffa0f00c288] vdpau_chroma filter error: video color space matrix failure:
An invalid pointer was provided. Typically, this means that a NULL pointer was
provided for an 'output' parameter.
[0x7ffa0f00c288] vdpau_chroma filter debug: using video mixer 7


2/ it regularly spits "vo messages like this :

[0x7ffa14001248] vdpau_display vout display debug: surface status: 2


3/ regularly, it spits "warning: picture is too late to be displayed (missing
113 ms)" messages

4/ regularly stalls with messages like this :

[CODE][0x7f08e0c1f3f8] main input error: ES_OUT_SET_(GROUP_)PCR  is called too
late (pts_delay increased to 1000 ms
[0x7f08e0c1f3f8] main input error: ES_OUT_RESET_PCR called
[0x7f08a0001248] vdpau_display vout display debug: surface status: 2
[0x7f08d91e7188] main decoder warning: early picture skipped[/CODE]


On the same system, mplayer from 2013/08/19 plays the same file, using VDPAU
and UVD (checked watching the CPU usage) fine.

I opened a bug on the VLC bug tracker
(https://trac.videolan.org/vlc/ticket/9281) but according to them, the problem
seems to come from the Galllium VDPAU state tracker as they should not get
"surface status : 2" at this stage of their code (they obviously tested their
VDPAU implementation using Nvidia binary driver, which does not use the VDPAU
Gallium state tracker).
Mplayer doesn't seem to have problem here as it seems to not even try to
support 
CSC matrix generation or decoder capabilities.

I'm sorry if this bug report is a bit of a mess. There may be several issues
involved here, and some may be more VLC-related that mesa, but I don't know at
all how to sort them out. 
The VLC devs at least seem to consider that the "surface status : 2" and CSC
matrix generation problem are Gallium-related and need to be fixed before they
can look into the problem if it's still there.

-- 
You are receiving this mail because:
You are the assignee for the bug.