[Libva] bug with vpp commit

Zhao Yakui yakui.zhao at intel.com
Thu Aug 13 21:15:32 PDT 2015 

On 08/12/2015 04:42 PM, Zhao Yakui wrote:
> On 08/12/2015 04:35 PM, Peter Frühberger wrote:
>> What's the status here?
>>
>> @Zhao Yukui:
>> gst-launch-0.10 filesrc location="some-file.ts" ! tsdemux !
>> vaapidecode ! vaapipostproc deinterlace=1 deinterlace-method=3 !
>> vaapisink
>>
>> should also not work for you after the pointed out patch. Can you
>> verify this?
>
> Thanks for your info.
> In the past few days I struggled to build/install xbmc on my system. But
> it fails on my side.
>
> If the gst can also reproduce the issue, I will check it and see whether
> it can be reproduced.

Hi, Peter/Zaverel

     Yesterday I tried to install the kodi on ubuntu. And the issue is 
reproduced.
     Will you please try the attached two patches and see whether the 
issue is resolved?

Thanks.
      Yakui

>
> Thanks.
>
>>
>> And in the mean time is it possible to revert this patch? I am afraid
>> that it will end in a support nightmare, when that one reaches a
>> stable branch.
>>
>> Thanks much
>> Peter
>>
>> 2015-08-01 12:07 GMT+02:00 zaverel<zaverel at free.fr>:
>>> Hello,
>>>
>>> I del and recompile
>>>
>>> - i965_vpp_clear_surface(ctx,&proc_context->pp_context,
>>> - obj_surface,
>>> - pipeline_param->output_background_color);
>>>
>>>
>>> but no change.
>>>
>>> libva.log
>>> http://www.filehosting.org/file/details/501119/libva.log.38445.7fe90c0a20f0
>>>
>>>
>>>
>>> See you.
>>>
>>>
>>> Le 31/07/2015 11:46, Zhao Yakui a écrit :
>>>>
>>>> On 07/29/2015 11:41 PM, zaverel wrote:
>>>>>
>>>>> Yes,
>>>>> if i unpatch :
>>>>> VPP: Combine the CSC/Scaling together to optimize the performance
>>>>>
>>>>> kodi / softhddevice work.
>>>>>
>>>>> I try all patch (one by one )with clean git.
>>>>>
>>>>>
>>>>
>>>> Hi, Zaverel
>>>>
>>>> I checked the libva log you uploaded.
>>>> In your test the deinterlace VPP is called. But the CSC/combine is
>>>> executed only after the VPP deinterlace is finished. And it has no
>>>> effect on
>>>> VPP deinterlace.
>>>> But it is quite strange that the XBMC is broken.
>>>> Will you please remove the function call of i965_vpp_clear_surface in
>>>> the corresponding commit and double check whether the issue exists?
>>>>
>>>> Thanks.
>>>> Yakui
>>>>
>>>>
>>>>> i upload libva.log (36 Mo) here:
>>>>>
>>>>>
>>>>> http://www.filehosting.org/file/details/500675/libva.log.61262.7f4634010c30
>>>>>
>>>>>
>>>>> See you.
>>>>>
>>>>>
>>>>> Le 29/07/2015 15:45, Zhao, Yakui a écrit :
>>>>>>
>>>>>> Hi, Peter/Zaverel
>>>>>> I am totally confused by the info.
>>>>>> From the info in previous email it seems that it can work after you
>>>>>> remove the background command.
>>>>>> Is what you mean to remove the i965_vpp_clear_surface in the
>>>>>> corresponding driver commit or revert the corresponding commit?
>>>>>> (Sorry that I am confused).
>>>>>>> Try to remove the background command and see if that works. The
>>>>>>>>>
>>>>>>>>> blackening can only come from that one
>>>>>>
>>>>>> In fact my original patch is only to combine the CSC/scale together
>>>>>> after it finishes all the VPP filter operation. It is the last step.
>>>>>> Not sure why the VPP in XBMC is affected.
>>>>>> I also try to download the XBMC. But I fail to build it as it
>>>>>> requires
>>>>>> the libmpeg2-devel package(But the libmpeg2-devel is deprecated on
>>>>>> my system. And there is no luck although the option of with-ffmpeg is
>>>>>> added). So it is difficult to reproduce the issue on my side.
>>>>>>
>>>>>> Will you please add one /etc/libva.conf like the following and send
>>>>>> out the libva trace log?
>>>>>>> LIBVA_TRACE=/pathto/tracelog
>>>>>>> LIBVA_TRACE_BUFDATA=1
>>>>>>
>>>>>> Thanks.
>>>>>> Yakui
>>>>>>
>>>>>> -----Original Message-----
>>>>>> From: Peter Frühberger [mailto:peter.fruehberger at gmail.com]
>>>>>> Sent: Wednesday, July 29, 2015 8:42 PM
>>>>>> To: zaverel
>>>>>> Cc: Zhao, Yakui; libva at lists.freedesktop.org
>>>>>> Subject: Re: [Libva] bug with vpp commit
>>>>>>
>>>>>> That's most likely not related to the patch.
>>>>>>
>>>>>> It just means vdr is sending 1 forward reference, but using VPP-BOB.
>>>>>> It is ignored in driver and it only warns.
>>>>>>
>>>>>> @Zhao: For you it is working? With e.g. gstreamer and vaapi? and
>>>>>> deinterlace-method>1?
>>>>>>
>>>>>> Best regards
>>>>>> Peter
>>>>>>
>>>>>> 2015-07-29 14:39 GMT+02:00 zaverel<zaverel at free.fr>:
>>>>>>>
>>>>>>> Hello,
>>>>>>>
>>>>>>> I just try your patch but no change, that's not work.
>>>>>>>
>>>>>>>
>>>>>>> Not sur if usefull , but in softhddevice log ,
>>>>>>>
>>>>>>>
>>>>>>> Jul 29 14:02:35 [vdr] vaapi/vpp: Wrong number of forward references.
>>>>>>> Needed 0, got 1 Jul 29 14:02:35 [vdr] Changing to 0 forward
>>>>>>> reference
>>>>>>> surfaces for postprocessing_
>>>>>>>
>>>>>>> See you.
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Le 29/07/2015 02:48, Zhao Yakui a écrit :
>>>>>>>>
>>>>>>>> On 07/28/2015 10:18 PM, Peter Frühberger wrote:
>>>>>>>>>
>>>>>>>>> Yeah :-)
>>>>>>>>>
>>>>>>>>> Try to remove the background command and see if that works. The
>>>>>>>>> blackening can only come from that one.
>>>>>>>>>
>>>>>>>>> As it breaks working applications - it should be fixed by the
>>>>>>>>> original commiter.'
>>>>>>>>
>>>>>>>>
>>>>>>>> Hi, Peter/Zaverel
>>>>>>>>
>>>>>>>> Will you please try the attached patch and see whether the issue
>>>>>>>> is resovled ?
>>>>>>>>
>>>>>>>> Thanks
>>>>>>>> Yakui
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Best regards
>>>>>>>>> Peter
>>>>>>>>>
>>>>>>>>> 2015-07-28 16:09 GMT+02:00 zaverel<zaverel at free.fr>:
>>>>>>>>>>
>>>>>>>>>> I tried both method , recompile and none work.
>>>>>>>>>>
>>>>>>>>>> I do another test but with
>>>>>>>>>> https://github.com/pesintta/vdr-plugin-softhddevice from
>>>>>>>>>> vdrportal.de
>>>>>>>>>>
>>>>>>>>>> "A software and GPU emulated HD output device plugin for VDR
>>>>>>>>>> (with
>>>>>>>>>> VA-API VPP additions) "
>>>>>>>>>>
>>>>>>>>>> And it does not work like for kodi.
>>>>>>>>>> Same bug with this commit.
>>>>>>>>>>
>>>>>>>>>> See you.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Le 28/07/2015 14:55, Peter Frühberger a écrit :
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> i965pp_context->filter_flags = VA_FILTER_SCALING_HQ;
>>>>>>>>>>>
>>>>>>>>>>> This does not consider any other flags already set - if
>>>>>>>>>>> VA_FILTER_SCALING_HQ is needed it should be ored (|=) onto the
>>>>>>>>>>> flags field as without it would loose everything else set.
>>>>>>>>>>>
>>>>>>>>>>> Is that the problem? Can you just remove the above mentioned
>>>>>>>>>>> line
>>>>>>>>>>> - that makes it working again? And as second test, try:
>>>>>>>>>>>
>>>>>>>>>>> i965pp_context->filter_flags |= VA_FILTER_SCALING_HQ; instead?
>>>>>>>>>>>
>>>>>>>>>>> Best regards
>>>>>>>>>>> Peter
>>>>>>>>>>>
>>>>>>>>>>> 2015-07-28 14:41 GMT+02:00 zaverel<zaverel at free.fr>:
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Hello,
>>>>>>>>>>>>
>>>>>>>>>>>> pvr mean personnal video recorder , it's for live tv.
>>>>>>>>>>>> Bug seem only apply to pvr (live tv) and not for regular video.
>>>>>>>>>>>>
>>>>>>>>>>>> when i check for VPP ,i don't realise testing with old divx is
>>>>>>>>>>>> not a good idea.
>>>>>>>>>>>> That why i thinked losing VPP capability.
>>>>>>>>>>>> My bad.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> So with commit
>>>>>>>>>>>> VPP: Combine the CSC/Scaling together to optimize the
>>>>>>>>>>>> performance
>>>>>>>>>>>> (2015-07-15)
>>>>>>>>>>>>
>>>>>>>>>>>> kodi have any vaapi deinterlace working but just for pvr.
>>>>>>>>>>>> If i unpatch that commit ,pvr work like before.
>>>>>>>>>>>>
>>>>>>>>>>>> I do a report http://trac.kodi.tv/ticket/16161
>>>>>>>>>>>>
>>>>>>>>>>>> See you.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Le 28/07/2015 02:20, Zhao Yakui a écrit :
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 07/28/2015 03:48 AM, zaverel wrote:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Hello,
>>>>>>>>>>>>>>
>>>>>>>>>>>>> Hi, Zaverel
>>>>>>>>>>>>>
>>>>>>>>>>>>> Sorry that I am confused by the issue.
>>>>>>>>>>>>> What is the PVR? What is the meaning of losing VPP capability?
>>>>>>>>>>>>>
>>>>>>>>>>>>> Thanks.
>>>>>>>>>>>>> Yakui
>>>>>>>>>>>>>
>>>>>>>>>>>>>> With some more test
>>>>>>>>>>>>>> there is no bug with video but for pvr.
>>>>>>>>>>>>>> Only deinterlace: auto work.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> And yes VPP is enumerated. Sorry.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I will report on trac.kodi.tv later if need.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> link to kodi.log
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> http://xbmclogs.com/pdjocrcyo
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> See you.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Le 27/07/2015 20:58, Peter Frühberger a écrit :
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> That's wrong.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> VPP is enumerated and we use what the driver gives us. A
>>>>>>>>>>>>>>> kodi
>>>>>>>>>>>>>>> debuglog wiill tell that.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> So, post one to trac.kodi.tv if you feel that this is a kodi
>>>>>>>>>>>>>>> bug.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Best regards
>>>>>>>>>>>>>>> Peter
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> 2015-07-27 20:56 GMT+02:00 zaverel<zaverel at free.fr>:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Hello,
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Whatever resolution.
>>>>>>>>>>>>>>>> In fact there is black screen because kodi try to do vpp
>>>>>>>>>>>>>>>> but
>>>>>>>>>>>>>>>> since vpp capability is lost , that didn't work.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> If i switch another available deinterlace ,video is back.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> See you.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On 07/25/2015 06:09 PM, zaverel wrote:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> hello,
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> i use latest libva and libva-intel-driver master
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> with commit : VPP: Combine the CSC/Scaling together to
>>>>>>>>>>>>>>>> optimize the performance
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> kodi don't work anymore with any vpp : black screen and
>>>>>>>>>>>>>>>> only
>>>>>>>>>>>>>>>> refresh overlay skin (no video).
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> So i revert the commit and that work again.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> who 's the guilty ? libva-intel-driver or kodi ?
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Hi, Zaverel
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Thanks for reporting the issue.
>>>>>>>>>>>>>>>> I will try the kodi and see whether the issue can be
>>>>>>>>>>>>>>>> reproduced.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> BTW: Which resolution is used when reproducing the issue?
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Thanks.
>>>>>>>>>>>>>>>> Yakui
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> See you.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Intel Core i3 3220T (Ivy Bridge hd graphics 2500)
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> lspci -vxxx -s 0:02.0
>>>>>>>>>>>>>>>> 00:02.0 VGA compatible controller: Intel Corporation Xeon
>>>>>>>>>>>>>>>> E3-1200 v2/3rd Gen Core processor Graphics Controller (rev
>>>>>>>>>>>>>>>> 09) (prog-if 00 [VGA
>>>>>>>>>>>>>>>> controller])
>>>>>>>>>>>>>>>> DeviceName: Onboard IGD
>>>>>>>>>>>>>>>> Subsystem: Micro-Star International Co., Ltd. [MSI] Xeon
>>>>>>>>>>>>>>>> E3-1200 v2/3rd Gen Core processor Graphics Controller
>>>>>>>>>>>>>>>> Flags: bus master, fast devsel, latency 0, IRQ 25 Memory at
>>>>>>>>>>>>>>>> f7800000 (64-bit, non-prefetchable) [size=4M] Memory at
>>>>>>>>>>>>>>>> e0000000 (64-bit, prefetchable) [size=256M] I/O ports at
>>>>>>>>>>>>>>>> f000
>>>>>>>>>>>>>>>> [size=64] Expansion ROM at<unassigned> [disabled]
>>>>>>>>>>>>>>>> Capabilities: [90] MSI: Enable+ Count=1/1 Maskable- 64bit-
>>>>>>>>>>>>>>>> Capabilities: [d0] Power Management version 2
>>>>>>>>>>>>>>>> Capabilities: [a4] PCI Advanced Features Kernel driver in
>>>>>>>>>>>>>>>> use: i915 Kernel modules: i915
>>>>>>>>>>>>>>>> 00: 86 80 52 01 07 04 90 00 09 00 00 03 00 00 00 00
>>>>>>>>>>>>>>>> 10: 04 00 80 f7 00 00 00 00 0c 00 00 e0 00 00 00 00
>>>>>>>>>>>>>>>> 20: 01 f0 00 00 00 00 00 00 00 00 00 00 62 14 11 21
>>>>>>>>>>>>>>>> 30: 00 00 00 00 90 00 00 00 00 00 00 00 0b 01 00 00
>>>>>>>>>>>>>>>> 40: 09 00 0c 01 92 80 80 e2 d0 00 50 14 00 00 00 00
>>>>>>>>>>>>>>>> 50: 41 02 00 00 11 00 00 00 00 00 00 00 01 00 a0 cf
>>>>>>>>>>>>>>>> 60: 00 00 02 00 00 00 00 00 00 00 00 00 00 00 00 00
>>>>>>>>>>>>>>>> 70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>>>>>>>>>>>>>>>> 80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>>>>>>>>>>>>>>>> 90: 05 d0 01 00 0c f0 e0 fe b1 41 00 00 00 00 00 00
>>>>>>>>>>>>>>>> a0: 00 00 00 00 13 00 06 03 00 00 00 00 00 00 00 00
>>>>>>>>>>>>>>>> b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>>>>>>>>>>>>>>>> c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>>>>>>>>>>>>>>>> d0: 01 a4 22 00 00 00 00 00 00 00 00 00 00 00 00 00
>>>>>>>>>>>>>>>> e0: 00 00 00 00 00 00 00 00 00 80 00 00 00 00 00 00
>>>>>>>>>>>>>>>> f0: 00 00 00 00 00 00 00 00 00 00 09 00 18 90 5a cd
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>>>>> Libva mailing list
>>>>>>>>>>>>>>>> Libva at lists.freedesktop.org
>>>>>>>>>>>>>>>> http://lists.freedesktop.org/mailman/listinfo/libva
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> _______________________________________________
>>>>>>>>>>>>>> Libva mailing list
>>>>>>>>>>>>>> Libva at lists.freedesktop.org
>>>>>>>>>>>>>> http://lists.freedesktop.org/mailman/listinfo/libva
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>
>>
>>
>>
>
> _______________________________________________
> Libva mailing list
> Libva at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/libva

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 0002-VPP-Initialize-the-uninitialed-surface-to-avoid-NULL.patch
URL: <http://lists.freedesktop.org/archives/libva/attachments/20150814/d94f3410/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 0001-VPP-Flush-batch-buffer-to-assure-that-previous-GPU-o.patch
URL: <http://lists.freedesktop.org/archives/libva/attachments/20150814/d94f3410/attachment-0001.ksh>

More information about the Libva mailing list