[Cin] QSV profile debuggin thread

[Andrew Randrianasulu]

сб, 12 окт. 2024 г., 00:36 Terje J. Hanssen <terjejhanssen at gmail.com>:

>
>
> Den 11.10.2024 23:13, skrev Andrew Randrianasulu:
>
>
>
> сб, 12 окт. 2024 г., 00:04 Terje J. Hanssen <terjejhanssen at gmail.com>:
>
>>
>>
>> Den 11.10.2024 22:39, skrev Andrew Randrianasulu:
>>
>>
>>
>> пт, 11 окт. 2024 г., 23:36 Terje J. Hanssen <terjejhanssen at gmail.com>:
>>
>>>
>>>
>>> Den 11.10.2024 22:13, skrev Terje J. Hanssen:
>>>
>>>
>>>
>>> Den 11.10.2024 21:04, skrev Andrew Randrianasulu:
>>>
>>> So, Terje, can you run your usual working ffmpeg command encoding av1
>>> via system fmmpeg/qsv but with added
>>>
>>> -loglevel debug
>>>
>>> ?
>>>
>>> and post full output here.
>>>
>>> Then we can set same loglevel in bin/ffmpeg/encode.opts
>>>
>>> so we hopefully will spot difference in parameters send down to line ...
>>>
>>> then we can try to make ffmpeg fail by adding say -profile=main to
>>> ffmpeg's command line and see more detailed error
>>>
>>> then hopefully we can tweak qsv_av1 profile back into usability.
>>>
>>> Repeat for vp9/h264.
>>>
>>>
>>>
>>> Andrew, I think use two input video examples
>>> 1) SD-DV
>>> 2) HDV
>>>
>>> Currently ffmpeg-7-7.0.2 on Tumbleweed-Slowroll (but have also
>>> ffmpeg-7-7.1 on Leap if wanted)
>>> I use output.log file. For some reason the terminal output alone is
>>> seemingly more verbose, and longer than I can scroll back to copy.
>>>
>>> Oh, the reason was the ffmpeg output file existed and it wasn't
>>> overwritten.
>>> I also shows that the output is too long for Cingg messages, and I try
>>> to post the two log files compressed by email if acceptable(?)
>>>
>>
>> I think only parts where we see setup.of MXF session are important, but I
>> hope log files compress well. If they still too big post them to me
>> directly.
>>
>>
>> I posted them directly this time  as .gz
>> If you tell me what to 'grep' or search for I can extract it and post it
>> here next (searched 'mxf without finding something) .
>>
>
>
> true, it was prefixed by
>
>
> [av1_qsv @
>
> try to search for av1_qsv lines ?
>
> Anyway, I think we get our working baseline, now (or may be tomorrow?
> depend on localtime) it will be time to get creative with ffmpeg switches.
>
> You probably can save lig files in folders named
>
>
> "lig files" (similar for h264/hevc/vp9 ?)
>


log files - typo!

>
>
> working
> param1_level10_profile_main
> param2_level30_profile_main10
> param3_level0_profile_high_notworking
>
> so it will be easer to see where working vs non-working logs are.
>
> Quote a chore. I suggest to create just few working and few non-working
> cases.
>
> May be for tommorow?
>
>
> Late now, so good suggestion.
> Tomorrow I  will be some on and off.
>

I tend to stay up at night, but then I'll sleep most of light day!
(depending on my dog of course).

so .. see you!




>
>
>
>
>>
>>
>>
>>> 1)
>>> 576i SD-DV -> AV1_QSV at bv1570k
>>> --------------------------------
>>> ffmpeg -loglevel debug -hwaccel_output_format qsv -qsv_device
>>> /dev/dri/renderD128 -i dv01_07.dv -c:v av1_qsv -preset 4 -b:v 1570k
>>> dv01_07_av1_qsv_pr4_bv1570k.webm > dv01_07_av1_qsv.log 2>&1
>>>
>>> 2)
>>> 1080i HDV->AV1_QSV at bv3700k
>>> -----------------------------
>>> fmpeg -hide_banner -hwaccel_output_format qsv -qsv_device
>>> /dev/dri/renderD128 -extra_hw_frames 16 -i hdv09_04.m2t -c:v av1_qsv
>>> -preset 4 -b:v 3700k hdv09_04_av1_qsv_pr4_bv3700k.webm >
>>> hdv09_04_av1_qsv.log 2>&1
>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cinelerra-gg.org/pipermail/cin/attachments/20241012/4a40a253/attachment-0001.htm>