[Nouveau] VDPAU DEINTERLACE
poma
pomidorabelisima at gmail.com
Mon May 9 18:55:07 UTC 2016
On 09.05.2016 20:45, Ilia Mirkin wrote:
> You can try playing with pstate in /sys/kernel/debug/dri/0/pstate
>
# cat /sys/kernel/debug/dri/0/pstate
0f: core 567 MHz shader 1400 MHz memory 400 MHz
AC: core 566 MHz shader 1400 MHz memory 399 MHz
±1 MHz :)
> On Mon, May 9, 2016 at 2:42 PM, poma <pomidorabelisima at gmail.com> wrote:
>> On 09.05.2016 19:37, Ilia Mirkin wrote:
>>> Mesa only supports the non-spatial temporal deinterlace (deint=3). I'm
>>> guessing that due to some unfortunate issues, you're no longer getting
>>> hw accelerated video decoding. Check in vdpauinfo to make sure that
>>> it's indeed showing the relevant codec as supported. If not, you can
>>> turn that back on by updating to mesa 11.2.2, or downgrading your
>>> kernel to 4.2 or earlier. (The issue only affects G98 and MCP77/MCP79
>>> IGPs.)
>>>
>>
>> With the -Mplayer- vdpau decoding works, at least with the -progressive- scan type,
>> -interlaced- scan type (DVBT-576i/1080i) is questionable,
>> especially when runs within vlc or xine, even without vdpau deinterlacer,
>> Xorg crash dump, satisfaction guarantee.
>>
>>
>> $ vdpauinfo
>> display: :0.0 screen: 0
>> API version: 1
>> Information string: G3DVL VDPAU Driver Shared Library version 1.0
>> ...
>>
>> Decoder capabilities:
>>
>> name level macbs width height
>> ----------------------------------------------------
>> MPEG1 0 16384 2048 2048
>> MPEG2_SIMPLE 3 16384 2048 2048
>> MPEG2_MAIN 3 16384 2048 2048
>> H264_BASELINE 41 16384 2048 2048
>> H264_MAIN 41 16384 2048 2048
>> H264_HIGH 41 16384 2048 2048
>> VC1_SIMPLE 1 16384 2048 2048
>> VC1_MAIN 2 16384 2048 2048
>> VC1_ADVANCED 4 16384 2048 2048
>> MPEG4_PART2_SP --- not supported ---
>> ...
>>
>> Video mixer:
>>
>> feature name sup
>> ------------------------------------
>> DEINTERLACE_TEMPORAL y
>> DEINTERLACE_TEMPORAL_SPATIAL -
>> INVERSE_TELECINE -
>> NOISE_REDUCTION y
>> SHARPNESS y
>> LUMA_KEY -
>> ...
>>
>>> If you are, in fact, getting hw video decoding acceleration, then it
>>> could be that your GPU is clocked too low. You could attempt
>>> reclocking to a higher pstate and seeing what happens.
>>>
>>
>> # nvclock --speeds
>> ...
>> Memory clock: 399.600 MHz
>> GPU clock: 612.000 MHz
>>
>> # nvclock --info
>> ...
>> Performance level 0: gpu 567MHz/shader 1400MHz/memory 400MHz/100%
>>
>> $ dmesg -t | grep pstate
>> ...
>> Kernel command line: ... nouveau.pstate=1 ...
>> nouveau: unknown parameter 'pstate' ignored
>> -4.5.2-
>>
>> Is there a room for reinforcement, or
>> NVIDIA G98 DEINTERLACER: ability without capability, i.e. underpowered GPU?
>>
>>>
>>> On Thu, May 5, 2016 at 1:12 AM, poma <pomidorabelisima at gmail.com> wrote:
>>>>
>>>> NVIDIA G98
>>>> mesa-dri-drivers-11.2.1-2.20160501.fc22.x86_64
>>>> (incl. mesa commit 38fcf7c)
>>>>
>>>>
>>>> vdpauinfo | grep -i deint
>>>> DEINTERLACE_TEMPORAL y
>>>> DEINTERLACE_TEMPORAL_SPATIAL -
>>>>
>>>>
>>>> https://cgit.freedesktop.org/vdpau/libvdpau/tree/include/vdpau/vdpau.h#n3420
>>>> #define VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL ((VdpVideoMixerFeature)0)
>>>> /**
>>>> * \hideinitializer
>>>> * \brief A VdpVideoMixerFeature.
>>>> *
>>>> * When requested and enabled, this enables a more advanced
>>>> * version of temporal de-interlacing, that additionally uses
>>>> * edge-guided spatial interpolation.
>>>> *
>>>> * When multiple de-interlacing options are requested and
>>>> * enabled, the back-end implementation chooses the best
>>>> * algorithm to apply.
>>>> */
>>>> #define VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL_SPATIAL ((VdpVideoMixerFeature)1)
>>>> /**
>>>> * \hideinitializer
>>>> * \brief A VdpVideoMixerFeature.
>>>> *
>>>> * When requested and enabled, cadence detection will be enabled
>>>> * on interlaced content and the video mixer will try to extract
>>>> * progressive frames from pull-down material.
>>>> */
>>>>
>>>>
>>>> https://cgit.freedesktop.org/vdpau/libvdpau/tree/include/vdpau/vdpau.h#n606
>>>> * \subsection deint_adv Advanced De-interlacing
>>>> *
>>>> * Operation of both temporal and temporal-spatial de-interlacing is
>>>> * identical; the only difference is the internal processing the algorithm
>>>> * performs in generating the output frame.
>>>> *
>>>>
>>>>
>>>> man 1 mplayer
>>>> ...
>>>> vdpau (X11 only)
>>>> ...
>>>> deint=<-4-4>
>>>> ...
>>>> Select deinterlacing mode (default: -3). Positive values
>>>> choose mode and enable deinterlacing. Corresponding nega‐
>>>> tive values select the same deinterlacing mode, but do
>>>> not enable deinterlacing on startup (useful in configura‐
>>>> tion files to specify what mode will be enabled by the
>>>> "D" key). All modes respect --field-dominance.
>>>>
>>>> 0 same as -3
>>>>
>>>> 1 Show only first field, similar to --vf=field.
>>>>
>>>> 2 Bob deinterlacing, similar to --vf=tfields=1.
>>>>
>>>> 3 motion adaptive temporal deinterlacing. May lead
>>>> to A/V desync with slow video hardware and/or high
>>>> resolution.
>>>>
>>>> 4 motion adaptive temporal deinterlacing with
>>>> edge-guided spatial interpolation. Needs fast
>>>> video hardware.
>>>>
>>>>
>>>> Reading all this, am I correctly concluded,
>>>> what is supported within NVIDIA G98 HW is DEINTERLACE_TEMPORAL,
>>>> which should be engaged with Mplayer's 'vdpau:deint=4' option?
>>>>
>>>> Then again, what DEINTERLACE_TEMPORAL_SPATIAL represents?
>>>> As reading the 'vdpauinfo' output it should not be supported.
>>>> Is it associated with Mplayer's 'vdpau:deint=3' option,
>>>> which in turn works, so to speak?
>>>>
>>>> mplayer -vo vdpau:deint=[34] -vc ffmpeg12vdpau dvb://2@DVBT
>>>>
>>>> Although they achieve solid deinterlacing result,
>>>> vdpau:deint=3 and vdpau:deint=4 tend to produce:
>>>>
>>>> ************************************************
>>>> **** Your system is too SLOW to play this! ****
>>>> ************************************************
>>>>
>>>> Rest of the deinterlacing modes - 1 and 2, are not so great.
>>>>
>>>>
>>>> _______________________________________________
>>>> Nouveau mailing list
>>>> Nouveau at lists.freedesktop.org
>>>> https://lists.freedesktop.org/mailman/listinfo/nouveau
>>
More information about the Nouveau
mailing list