[RESOLVED] DGMVCSource b22 and VC++2012

This is the home of QuickSync (aka Intel Media SDK) stuff.
DAE avatar
Nico83500
Posts: 78
Joined: Thu Jan 02, 2014 9:28 am

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by Nico83500 »

Ok thanks ;-)
I can confirm movies are bug free with the b23 on my system.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Hi Donald. Sorry to revive this old thread, but I've tried to update libmfxsw32.dll to the latest version, and DGMVCSource seems to produce only black frames with some BDs. It takes a long time before decoding the first frame, and then all frames are black (and the "decoding" is extremely fast!)

The version of libmfxsw32 I've tried is v6.15.6.2, dated 23 sep 2015 and taken from INDE 2015 update 2.1 (NOT update 2 RTM) "w_inde_2015.2.048.exe". The last version that worked fine with your plugin (and FRIMSource) was v6.14.11.28. Since I don't have the right CPU, I can't test the library in hardware mode. Personally, I'm happy with v6.14.11.28, but I suppose that the latest version may be better, or even necessary for correct operation in hardware mode with new CPUs, and therefore I would like to update it in the distribution of BD3D2MK3D.

Strangely v6.15.6.2 works fine with FRIMSource (without update). Therefore, I guess that there is probably only a little thing to do to make DGMVCSource compatible with the latest version. Perhaps just a new compilation? Note also that the latest version works fine with many BD3Ds, so there might be an incompatibility occurring only with some BDs.

Here is a sample: http://download.videohelp.com/r0lZ/tmp/ ... .15.6.2.7z

Any chance to have an update?

Thanks in advance!

[EDIT]
Just tried your plugin with libmfxsw32 v6.15.3.12 (from INDE update 2 RTM, 22 apr 2015) and it works fine. So perhaps the black frames are caused by a bug in the latest Intel library. Anyway, I will use v6.15.3.12 for BD3D2MK3D, at least while there is no new version of your plugin or of the libmfxsw32 DLL. Therefore, a fix is less urgent, but you may want to consider it anyway.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

I'll look into it.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

I installed update 2.1 and find that DGDecIM still works fine. But I duplicated your issue with DGMVCDec and your stream.

Can you please give me a sample set that works with 2.1? Then I can compare what happens with the two in the debugger. Thanks.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

The library appears to be under-reporting the number of surfaces required for decoding. If I allocate the suggested amount plus 4, then the software decode works. But this workaround does not fix HW decoding. Does FRIM do only software decode? If not, I wonder if it too is failing in HW mode.

I'll tinker some more with HW decoding. If I can't get a workaround for that would you still want the SW workaround?
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Well, I don't know. If you're sure it's a bug in the Intel lib, I suppose it will be fixed soon, and I can skip this version. However, if the fix is simple to implement and has no consequence with the previous correct and forthcoming versions, maybe it is worth doing it, even if the fix works only in SW mode. But I suppose that I will have to distribute the last version that works in both modes anyway. So, I don't need the fix, and it's up to you.

BTW, I wonder why DGDecIM has no problem. Do you allocate anyway more surfaces than theoretically required? If it's the case, why don't you do the same thing with DGMVCDecode?

FRIM does hw decoding too, but I don't know if it fails with v6.15.6.2 in hw mode. AFAIK, nobody has reported problems with that version, in sw or hw mode, with FRIM, DGMVCDecode or DGDecIM. I suppose that the 3DBD that has caused the problem is a rare case and/or most peoples use an older Intel library.

Can you please notify the bug to the guys at Intel? You will be able to explain the problem much better than me. (Of course, you can give them the link to the sample if you wish.)

Thanks anyway for your investigation.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

r0lZ wrote: BTW, I wonder why DGDecIM has no problem. Do you allocate anyway more surfaces than theoretically required? If it's the case, why don't you do the same thing with DGMVCDecode?
It happens only for MVC decoding.
Can you please notify the bug to the guys at Intel? You will be able to explain the problem much better than me. (Of course, you can give them the link to the sample if you wish.)
Based on my past experience with Intel, it would just be a waste of my time.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

OK, thanks anyway.

Will you do an update or should I still distribute the current version of DGMVCDecode with BD3D2MK3D ?
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Let's stick with the current version for now. If you want to include the v6.15.3.12 DLL that should be OK.

I'm not aware of anything important for 3D stuff being fixed in the latest INDE.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

OK. I hope that the next version of the Intel lib will be bug free. I prefer to be up to date anyway. But you're right. It's not urgent, and the MVC decoder works perfectly with older versions.

Thanks again.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Just to let you know, I've just tested the new libmfxsw32 v7.15.10.28 (from the intel_media_sdk_2016.msi) and it produces still the black frames with DGMVCSource.
Pity.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Hi Donald.

Statula (an user of BD3D2MK3D) has reported the problem of the black frames with GDMVCSource and a recent version of the Intel decoder installed with the driver for his hardware. This time, the BD causing problem is Jurassic World. So, it appears that indeed, the hw lib has the same bug than the sw lib, and that the problem originally noticed with Creature from the Black Lagoon happens also with other 3DBDs. However, Statula has also tested FRIMSource, and FRIMSource has no problem, even in hw mode. Also, Pistacho has replied to the thread that his MVCSource has no black frames problems. (I haven't tested it.) Therefore, it seems that there is a way to fix the bug, even when the latest hardware libs are used.

So, I come back with this. Can you try to fix the bug or implement a workaround? Perhaps it is sufficient to increase the number of surfaces? You wrote previously that that will not be sufficient in hw mode, but there must be a way to use the new libmfxhw32.dll without problem, since two decoders can do it. I can't help you, but I'm sure the author of FRIMSource will accept to help, if you need some additional info.

Anyway, thanks for considering my request.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

You could be right, of course. Resolving it would be a nontrivial project for me and the other authors do not in fact share their methods (I have tried in the past). But the main thing for me is I have no motivation about it. I'm mostly doing physics research these days, and Intel support is essentially non-existent. So let me ask you, before I devote any time to it, why is it so important to have DGMVCSource, when you have FRIMSource and MVCSource? It seems to me that it would be easier to just incorporate anything useful I have done feature-wise in DGMVCSource into FRIMSource or MVCSource.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Well, as you know, Pistacho's MVCSource cannot be distributed with BD3D2MK3D (or any other software) and doesn't work without installing his BDtoAVCHD. I don't want to force the users to install that ad supported program to use mine. Therefore, MVCSource is not and will never be an option.

FRIMSource is free and already distributed with BD3D2MK3D, but several peoples have reported problems in the past. I have never been able to verify them, but your program has always worked correctly for everybody, and I have used yours almost exclusively. It is still the default MVC decoder for BD3D2MK3D, because I know and trust it. It's a pity that the recent Intel libs do not work as expected.

Of course, I can remove your plugin from the BD3D2MK3D distribution and use only FRIM, but since it is possible that FRIM has other bugs that your plugin doesn't have, it would be a pity. Keeping both is a good safeguard in case of new problems (for example due to a new bug in a forthcoming version of the intel libs.)

I think also that it is better for you if your plugin works well in all circumstances, regardless of its usage in my program. But that's up to you. BTW, you wrote previously "I installed update 2.1 and find that DGDecIM still works fine. But I duplicated your issue with DGMVCDec and your stream." So, I suppose that the solution is already in the code of DGDecIM and that it should not be so difficult to implement it in DGMVCDecode (although I understand that finding what subtle difference is the cause of the bug can take a long time).

I know that Pistacho will probably refuse to share his code or even explain his method. He has always taken the programs and knowledge of the other programmers and never shared his sources or knowledge. But I think that videofan3d will accept if he has some time to spend on this. I can't be sure however. I can ask him if you wish.

Anyway, I don't want to force you to do it if you have other things to do or if you're not interested in that plugin any more. I will accept your decision, but I hope you'll devote some time to try anyway.

Thanks anyway for what you did, and what you will do! ;-)
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Fair enough. When I get my current physics paper revised and re-submitted I will revisit this stuff.

Can you give me samples for the Jurassic Park case? Thank you.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Huge thanks.

I don't have the Jurassic Park BD and can't give you a sample, but the sample of Creature from the Black Lagoon is still available here. The problem should be identical with Creature and Jurassic World. But let me know if you really want a sample from Jurassic World. I can ask it to Statula, but that may be difficult if he has no simple way to cut or upload it.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

OK, I'll get by for now with the first sample.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Happy New Year, r0lZ and everybody!

Possibly I have fixed the black frame problem. The performance may also be improved. Please try b24 and let me know the result. I have tested both SW and HW mode.

http://rationalqm.us/dgmvcsource/dgmvcsource100b24.zip

If it is OK, I will release properly with the source code.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Thanks, and happy new year to you too.
I'll test the beta immediately and let you know...
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

I did some tests in software mode, with Creature from the Black Lagoon and Everest, and the beta seems to work perfectly. :-)

I can't test myself the hardware mode, but I've asked the BD3D2MK3D users to test it in hw mode. That may take some time, though. I think that you can probably already release the final version immediately if you wish. I'm not in a hurry, so take your time anyway.

Huge thanks!
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Excellent. Thank you for your test results.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

And it has passed the test in hardware mode too, as you can see here.
Thanks again!
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Sweet. Rock on!
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

BTW, the b24 up there now is the release so go ahead and use that one with your great application r0lZ.
Post Reply